Showing 1–50 of 120 results for author: Ramchandran, K

Toward a Theory of Tokenization in LLMs

Authors: Nived Rajaraman, Jiantao Jiao, Kannan Ramchandran

Abstract: While there has been a large body of research attempting to circumvent tokenization for language modeling (Clark et al., 2022; Xue et al., 2022), the current consensus is that it is a necessary initial step for designing state-of-the-art performant language models. In this paper, we investigate tokenization from a theoretical point of view by studying the behavior of transformers on simple data ge… ▽ More While there has been a large body of research attempting to circumvent tokenization for language modeling (Clark et al., 2022; Xue et al., 2022), the current consensus is that it is a necessary initial step for designing state-of-the-art performant language models. In this paper, we investigate tokenization from a theoretical point of view by studying the behavior of transformers on simple data generating processes. When trained on data drawn from certain simple $k^{\text{th}}$-order Markov processes for $k > 1$, transformers exhibit a surprising phenomenon - in the absence of tokenization, they empirically fail to learn the right distribution and predict characters according to a unigram model (Makkuva et al., 2024). With the addition of tokenization, however, we empirically observe that transformers break through this barrier and are able to model the probabilities of sequences drawn from the source near-optimally, achieving small cross-entropy loss. With this observation as starting point, we study the end-to-end cross-entropy loss achieved by transformers with and without tokenization. With the appropriate tokenization, we show that even the simplest unigram models (over tokens) learnt by transformers are able to model the probability of sequences drawn from $k^{\text{th}}$-order Markov sources near optimally. Our analysis provides a justification for the use of tokenization in practice through studying the behavior of transformers on Markovian data. △ Less

Submitted 12 April, 2024; originally announced April 2024.

Comments: 58 pages, 10 figures

Learning to Understand: Identifying Interactions via the Möbius Transform

Authors: Justin S. Kang, Yigit E. Erginbas, Landon Butler, Ramtin Pedarsani, Kannan Ramchandran

Abstract: One of the key challenges in machine learning is to find interpretable representations of learned functions. The Möbius transform is essential for this purpose, as its coefficients correspond to unique importance scores for sets of input variables. This transform is closely related to widely used game-theoretic notions of importance like the Shapley and Bhanzaf value, but it also captures crucial… ▽ More One of the key challenges in machine learning is to find interpretable representations of learned functions. The Möbius transform is essential for this purpose, as its coefficients correspond to unique importance scores for sets of input variables. This transform is closely related to widely used game-theoretic notions of importance like the Shapley and Bhanzaf value, but it also captures crucial higher-order interactions. Although computing the obius Transform of a function with $n$ inputs involves $2^n$ coefficients, it becomes tractable when the function is sparse and of low-degree as we show is the case for many real-world functions. Under these conditions, the complexity of the transform computation is significantly reduced. When there are $K$ non-zero coefficients, our algorithm recovers the Möbius transform in $O(Kn)$ samples and $O(Kn^2)$ time asymptotically under certain assumptions, the first non-adaptive algorithm to do so. We also uncover a surprising connection between group testing and the Möbius transform. For functions where all interactions involve at most $t$ inputs, we use group testing results to compute the Möbius transform with $O(Kt\log n)$ sample complexity and $O(K\mathrm{poly}(n))$ time. A robust version of this algorithm withstands noise and maintains this complexity. This marks the first $n$ sub-linear query complexity, noise-tolerant algorithm for the Möbius transform. In several examples, we observe that representations generated via sparse Möbius transform are up to twice as faithful to the original function, as compared to Shaply and Banzhaf values, while using the same number of terms. △ Less

Submitted 15 June, 2024; v1 submitted 4 February, 2024; originally announced February 2024.

Comments: 34 pages, 16 figures

Towards Optimal Statistical Watermarking

Authors: Baihe Huang, Hanlin Zhu, Banghua Zhu, Kannan Ramchandran, Michael I. Jordan, Jason D. Lee, Jiantao Jiao

Abstract: We study statistical watermarking by formulating it as a hypothesis testing problem, a general framework which subsumes all previous statistical watermarking methods. Key to our formulation is a coupling of the output tokens and the rejection region, realized by pseudo-random generators in practice, that allows non-trivial trade-offs between the Type I error and Type II error. We characterize the… ▽ More We study statistical watermarking by formulating it as a hypothesis testing problem, a general framework which subsumes all previous statistical watermarking methods. Key to our formulation is a coupling of the output tokens and the rejection region, realized by pseudo-random generators in practice, that allows non-trivial trade-offs between the Type I error and Type II error. We characterize the Uniformly Most Powerful (UMP) watermark in the general hypothesis testing setting and the minimax Type II error in the model-agnostic setting. In the common scenario where the output is a sequence of $n$ tokens, we establish nearly matching upper and lower bounds on the number of i.i.d. tokens required to guarantee small Type I and Type II errors. Our rate of $Θ(h^{-1} \log (1/h))$ with respect to the average entropy per token $h$ highlights potentials for improvement from the rate of $h^{-2}$ in the previous works. Moreover, we formulate the robust watermarking problem where the user is allowed to perform a class of perturbations on the generated texts, and characterize the optimal Type II error of robust UMP tests via a linear programming problem. To the best of our knowledge, this is the first systematic statistical treatment on the watermarking problem with near-optimal rates in the i.i.d. setting, which might be of interest for future works. △ Less

Submitted 6 February, 2024; v1 submitted 13 December, 2023; originally announced December 2023.

Pairwise Proximal Policy Optimization: Harnessing Relative Feedback for LLM Alignment

Authors: Tianhao Wu, Banghua Zhu, Ruoyu Zhang, Zhao** Wen, Kannan Ramchandran, Jiantao Jiao

Abstract: Large Language Models (LLMs) can acquire extensive world knowledge through pre-training on large corpora. However, due to exposure to low-quality data, LLMs may exhibit harmful behavior without aligning with human values. The dominant approach for steering LLMs towards beneficial behavior involves Reinforcement Learning with Human Feedback (RLHF), with Proximal Policy Optimization (PPO) serving as… ▽ More Large Language Models (LLMs) can acquire extensive world knowledge through pre-training on large corpora. However, due to exposure to low-quality data, LLMs may exhibit harmful behavior without aligning with human values. The dominant approach for steering LLMs towards beneficial behavior involves Reinforcement Learning with Human Feedback (RLHF), with Proximal Policy Optimization (PPO) serving as the default RL optimizer. Despite its effectiveness, PPO has limitations when optimizing rewards trained from comparison-based loss. Primarily, PPO is not invariant to equivalent reward functions containing identical preference information due to the need to calibrate the reward scale. Additionally, PPO's necessity for token-wise updates introduces complexity in both function approximation and algorithm design compared to trajectory-wise optimization. This paper proposes a new framework, reinforcement learning with relative feedback, and a novel trajectory-wise policy gradient algorithm, Pairwise Proximal Policy Optimization (P3O) that operates directly on comparative rewards. We show theoretically that P3O is invariant to equivalent rewards and avoids the complexity of PPO. Empirical evaluations demonstrate that P3O outperforms PPO in the KL-Reward trade-off and can align with human preferences as well as or better than prior methods. In summary, this work introduces a simpler yet effective approach for aligning LLMs to human preferences through relative feedback. △ Less

Submitted 9 October, 2023; v1 submitted 29 September, 2023; originally announced October 2023.

Comments: 19 pages, 5 figures

MRI Reconstruction with Side Information using Diffusion Models

Authors: Brett Levac, Ajil Jalal, Kannan Ramchandran, Jonathan I. Tamir

Abstract: Magnetic resonance imaging (MRI) exam protocols consist of multiple contrast-weighted images of the same anatomy to emphasize different tissue properties. Due to the long acquisition times required to collect fully sampled k-space measurements, it is common to only collect a fraction of k-space for each scan and subsequently solve independent inverse problems for each image contrast. Recently, the… ▽ More Magnetic resonance imaging (MRI) exam protocols consist of multiple contrast-weighted images of the same anatomy to emphasize different tissue properties. Due to the long acquisition times required to collect fully sampled k-space measurements, it is common to only collect a fraction of k-space for each scan and subsequently solve independent inverse problems for each image contrast. Recently, there has been a push to further accelerate MRI exams using data-driven priors, and generative models in particular, to regularize the ill-posed inverse problem of image reconstruction. These methods have shown promising improvements over classical methods. However, many of the approaches neglect the additional information present in a clinical MRI exam like the multi-contrast nature of the data and treat each scan as an independent reconstruction. In this work we show that by learning a joint Bayesian prior over multi-contrast data with a score-based generative model we are able to leverage the underlying structure between random variables related to a given imaging problem. This leads to an improvement in image reconstruction fidelity over generative models that rely only on a marginal prior over the image contrast of interest. △ Less

Submitted 6 June, 2023; v1 submitted 26 March, 2023; originally announced March 2023.

Greedy Pruning with Group Lasso Provably Generalizes for Matrix Sensing

Authors: Nived Rajaraman, Devvrit, Aryan Mokhtari, Kannan Ramchandran

Abstract: Pruning schemes have been widely used in practice to reduce the complexity of trained models with a massive number of parameters. In fact, several practical studies have shown that if a pruned model is fine-tuned with some gradient-based updates it generalizes well to new samples. Although the above pipeline, which we refer to as pruning + fine-tuning, has been extremely successful in lowering the… ▽ More Pruning schemes have been widely used in practice to reduce the complexity of trained models with a massive number of parameters. In fact, several practical studies have shown that if a pruned model is fine-tuned with some gradient-based updates it generalizes well to new samples. Although the above pipeline, which we refer to as pruning + fine-tuning, has been extremely successful in lowering the complexity of trained models, there is very little known about the theory behind this success. In this paper, we address this issue by investigating the pruning + fine-tuning framework on the overparameterized matrix sensing problem with the ground truth $U_\star \in \mathbb{R}^{d \times r}$ and the overparameterized model $U \in \mathbb{R}^{d \times k}$ with $k \gg r$. We study the approximate local minima of the mean square error, augmented with a smooth version of a group Lasso regularizer, $\sum_{i=1}^k \| U e_i \|_2$. In particular, we provably show that pruning all the columns below a certain explicit $\ell_2$-norm threshold results in a solution $U_{\text{prune}}$ which has the minimum number of columns $r$, yet close to the ground truth in training loss. Moreover, in the subsequent fine-tuning phase, gradient descent initialized at $U_{\text{prune}}$ converges at a linear rate to its limit. While our analysis provides insights into the role of regularization in pruning, we also show that running gradient descent in the absence of regularization results in models which {are not suitable for greedy pruning}, i.e., many columns could have their $\ell_2$ norm comparable to that of the maximum. To the best of our knowledge, our results provide the first rigorous insights on why greedy pruning + fine-tuning leads to smaller models which also generalize well. △ Less

Submitted 4 June, 2023; v1 submitted 20 March, 2023; originally announced March 2023.

Comments: 49 pages, 2 figures

Statistical Complexity and Optimal Algorithms for Non-linear Ridge Bandits

Authors: Nived Rajaraman, Yanjun Han, Jiantao Jiao, Kannan Ramchandran

Abstract: We consider the sequential decision-making problem where the mean outcome is a non-linear function of the chosen action. Compared with the linear model, two curious phenomena arise in non-linear models: first, in addition to the "learning phase" with a standard parametric rate for estimation or regret, there is an "burn-in period" with a fixed cost determined by the non-linear function; second, ac… ▽ More We consider the sequential decision-making problem where the mean outcome is a non-linear function of the chosen action. Compared with the linear model, two curious phenomena arise in non-linear models: first, in addition to the "learning phase" with a standard parametric rate for estimation or regret, there is an "burn-in period" with a fixed cost determined by the non-linear function; second, achieving the smallest burn-in cost requires new exploration algorithms. For a special family of non-linear functions named ridge functions in the literature, we derive upper and lower bounds on the optimal burn-in cost, and in addition, on the entire learning trajectory during the burn-in period via differential equations. In particular, a two-stage algorithm that first finds a good initial action and then treats the problem as locally linear is statistically optimal. In contrast, several classical algorithms, such as UCB and algorithms relying on regression oracles, are provably suboptimal. △ Less

Submitted 9 January, 2024; v1 submitted 12 February, 2023; originally announced February 2023.

Comments: Revised Section 3 and added an upper bound agnostic to the link function $f$

The Fair Value of Data Under Heterogeneous Privacy Constraints in Federated Learning

Authors: Justin Kang, Ramtin Pedarsani, Kannan Ramchandran

Abstract: Modern data aggregation often involves a platform collecting data from a network of users with various privacy options. Platforms must solve the problem of how to allocate incentives to users to convince them to share their data. This paper puts forth an idea for a \textit{fair} amount to compensate users for their data at a given privacy level based on an axiomatic definition of fairness, along t… ▽ More Modern data aggregation often involves a platform collecting data from a network of users with various privacy options. Platforms must solve the problem of how to allocate incentives to users to convince them to share their data. This paper puts forth an idea for a \textit{fair} amount to compensate users for their data at a given privacy level based on an axiomatic definition of fairness, along the lines of the celebrated Shapley value. To the best of our knowledge, these are the first fairness concepts for data that explicitly consider privacy constraints. We also formulate a heterogeneous federated learning problem for the platform with privacy level options for users. By studying this problem, we investigate the amount of compensation users receive under fair allocations with different privacy levels, amounts of data, and degrees of heterogeneity. We also discuss what happens when the platform is forced to design fair incentives. Under certain conditions we find that when privacy sensitivity is low, the platform will set incentives to ensure that it collects all the data with the lowest privacy options. When the privacy sensitivity is above a given threshold, the platform will provide no incentives to users. Between these two extremes, the platform will set the incentives so some fraction of the users chooses the higher privacy option and the others chooses the lower privacy option. △ Less

Submitted 4 February, 2024; v1 submitted 30 January, 2023; originally announced January 2023.

Comments: 29 pages, 5 figures, Accepted to TMLR

Efficiently Computing Sparse Fourier Transforms of $q$-ary Functions

Authors: Yigit Efe Erginbas, Justin Singh Kang, Amirali Aghazadeh, Kannan Ramchandran

Abstract: Fourier transformations of pseudo-Boolean functions are popular tools for analyzing functions of binary sequences. Real-world functions often have structures that manifest in a sparse Fourier transform, and previous works have shown that under the assumption of sparsity the transform can be computed efficiently. But what if we want to compute the Fourier transform of functions defined over a $q$-a… ▽ More Fourier transformations of pseudo-Boolean functions are popular tools for analyzing functions of binary sequences. Real-world functions often have structures that manifest in a sparse Fourier transform, and previous works have shown that under the assumption of sparsity the transform can be computed efficiently. But what if we want to compute the Fourier transform of functions defined over a $q$-ary alphabet? These types of functions arise naturally in many areas including biology. A typical workaround is to encode the $q$-ary sequence in binary, however, this approach is computationally inefficient and fundamentally incompatible with the existing sparse Fourier transform techniques. Herein, we develop a sparse Fourier transform algorithm specifically for $q$-ary functions of length $n$ sequences, dubbed $q$-SFT, which provably computes an $S$-sparse transform with vanishing error as $q^n \rightarrow \infty$ in $O(Sn)$ function evaluations and $O(S n^2 \log q)$ computations, where $S = q^{nδ}$ for some $δ< 1$. Under certain assumptions, we show that for fixed $q$, a robust version of $q$-SFT has a sample complexity of $O(Sn^2)$ and a computational complexity of $O(Sn^3)$ with the same asymptotic guarantees. We present numerical simulations on synthetic and real-world RNA data, demonstrating the scalability of $q$-SFT to massively high dimensional $q$-ary functions. △ Less

Submitted 15 January, 2023; originally announced January 2023.

Comments: 29 pages, 3 figures

Interactive Learning with Pricing for Optimal and Stable Allocations in Markets

Authors: Yigit Efe Erginbas, Soham Phade, Kannan Ramchandran

Abstract: Large-scale online recommendation systems must facilitate the allocation of a limited number of items among competing users while learning their preferences from user feedback. As a principled way of incorporating market constraints and user incentives in the design, we consider our objectives to be two-fold: maximal social welfare with minimal instability. To maximize social welfare, our proposed… ▽ More Large-scale online recommendation systems must facilitate the allocation of a limited number of items among competing users while learning their preferences from user feedback. As a principled way of incorporating market constraints and user incentives in the design, we consider our objectives to be two-fold: maximal social welfare with minimal instability. To maximize social welfare, our proposed framework enhances the quality of recommendations by exploring allocations that optimistically maximize the rewards. To minimize instability, a measure of users' incentives to deviate from recommended allocations, the algorithm prices the items based on a scheme derived from the Walrasian equilibria. Though it is known that these equilibria yield stable prices for markets with known user preferences, our approach accounts for the inherent uncertainty in the preferences and further ensures that the users accept their recommendations under offered prices. To the best of our knowledge, our approach is the first to integrate techniques from combinatorial bandits, optimal resource allocation, and collaborative filtering to obtain an algorithm that achieves sub-linear social welfare regret as well as sub-linear instability. Empirical studies on synthetic and real-world data also demonstrate the efficacy of our strategy compared to approaches that do not fully incorporate all these aspects. △ Less

Submitted 13 December, 2022; originally announced December 2022.

Comments: arXiv admin note: substantial text overlap with arXiv:2207.04143

Spectral Regularization Allows Data-frugal Learning over Combinatorial Spaces

Authors: Amirali Aghazadeh, Nived Rajaraman, Tony Tu, Kannan Ramchandran

Abstract: Data-driven machine learning models are being increasingly employed in several important inference problems in biology, chemistry, and physics which require learning over combinatorial spaces. Recent empirical evidence (see, e.g., [1], [2], [3]) suggests that regularizing the spectral representation of such models improves their generalization power when labeled data is scarce. However, despite th… ▽ More Data-driven machine learning models are being increasingly employed in several important inference problems in biology, chemistry, and physics which require learning over combinatorial spaces. Recent empirical evidence (see, e.g., [1], [2], [3]) suggests that regularizing the spectral representation of such models improves their generalization power when labeled data is scarce. However, despite these empirical studies, the theoretical underpinning of when and how spectral regularization enables improved generalization is poorly understood. In this paper, we focus on learning pseudo-Boolean functions and demonstrate that regularizing the empirical mean squared error by the L_1 norm of the spectral transform of the learned function reshapes the loss landscape and allows for data-frugal learning, under a restricted secant condition on the learner's empirical error measured against the ground truth function. Under a weaker quadratic growth condition, we show that stationary points which also approximately interpolate the training data points achieve statistically optimal generalization performance. Complementing our theory, we empirically demonstrate that running gradient descent on the regularized loss results in a better generalization performance compared to baseline algorithms in several data-scarce real-world problems. △ Less

Submitted 5 October, 2022; originally announced October 2022.

Interactive Recommendations for Optimal Allocations in Markets with Constraints

Authors: Yigit Efe Erginbas, Soham Phade, Kannan Ramchandran

Abstract: Recommendation systems when employed in markets play a dual role: they assist users in selecting their most desired items from a large pool and they help in allocating a limited number of items to the users who desire them the most. Despite the prevalence of capacity constraints on allocations in many real-world recommendation settings, a principled way of incorporating them in the design of these… ▽ More Recommendation systems when employed in markets play a dual role: they assist users in selecting their most desired items from a large pool and they help in allocating a limited number of items to the users who desire them the most. Despite the prevalence of capacity constraints on allocations in many real-world recommendation settings, a principled way of incorporating them in the design of these systems has been lacking. Motivated by this, we propose an interactive framework where the system provider can enhance the quality of recommendations to the users by opportunistically exploring allocations that maximize user rewards and respect the capacity constraints using appropriate pricing mechanisms. We model the problem as an instance of a low-rank combinatorial multi-armed bandit problem with selection constraints on the arms. We employ an integrated approach using techniques from collaborative filtering, combinatorial bandits, and optimal resource allocation to provide an algorithm that provably achieves sub-linear regret, namely $\tilde{\mathcal{O}} ( \sqrt{N M (N+M) RT} )$ in $T$ rounds for a problem with $N$ users, $M$ items and rank $R$ mean reward matrix. Empirical studies on synthetic and real-world data also demonstrate the effectiveness and performance of our approach. △ Less

Submitted 28 July, 2022; v1 submitted 8 July, 2022; originally announced July 2022.

Neurotoxin: Durable Backdoors in Federated Learning

Authors: Zhengming Zhang, Ashwinee Panda, Linyue Song, Yaoqing Yang, Michael W. Mahoney, Joseph E. Gonzalez, Kannan Ramchandran, Prateek Mittal

Abstract: Due to their decentralized nature, federated learning (FL) systems have an inherent vulnerability during their training to adversarial backdoor attacks. In this type of attack, the goal of the attacker is to use poisoned updates to implant so-called backdoors into the learned model such that, at test time, the model's outputs can be fixed to a given target for certain inputs. (As a simple toy exam… ▽ More Due to their decentralized nature, federated learning (FL) systems have an inherent vulnerability during their training to adversarial backdoor attacks. In this type of attack, the goal of the attacker is to use poisoned updates to implant so-called backdoors into the learned model such that, at test time, the model's outputs can be fixed to a given target for certain inputs. (As a simple toy example, if a user types "people from New York" into a mobile keyboard app that uses a backdoored next word prediction model, then the model could autocomplete the sentence to "people from New York are rude"). Prior work has shown that backdoors can be inserted into FL models, but these backdoors are often not durable, i.e., they do not remain in the model after the attacker stops uploading poisoned updates. Thus, since training typically continues progressively in production FL systems, an inserted backdoor may not survive until deployment. Here, we propose Neurotoxin, a simple one-line modification to existing backdoor attacks that acts by attacking parameters that are changed less in magnitude during training. We conduct an exhaustive evaluation across ten natural language processing and computer vision tasks, and we find that we can double the durability of state of the art backdoors. △ Less

Submitted 12 June, 2022; originally announced June 2022.

Comments: Appears in ICML 2022

Decentralized Competing Bandits in Non-Stationary Matching Markets

Authors: Avishek Ghosh, Abishek Sankararaman, Kannan Ramchandran, Tara Javidi, Arya Mazumdar

Abstract: Understanding complex dynamics of two-sided online matching markets, where the demand-side agents compete to match with the supply-side (arms), has recently received substantial interest. To that end, in this paper, we introduce the framework of decentralized two-sided matching market under non stationary (dynamic) environments. We adhere to the serial dictatorship setting, where the demand-side a… ▽ More Understanding complex dynamics of two-sided online matching markets, where the demand-side agents compete to match with the supply-side (arms), has recently received substantial interest. To that end, in this paper, we introduce the framework of decentralized two-sided matching market under non stationary (dynamic) environments. We adhere to the serial dictatorship setting, where the demand-side agents have unknown and different preferences over the supply-side (arms), but the arms have fixed and known preference over the agents. We propose and analyze a decentralized and asynchronous learning algorithm, namely Decentralized Non-stationary Competing Bandits (\texttt{DNCB}), where the agents play (restrictive) successive elimination type learning algorithms to learn their preference over the arms. The complexity in understanding such a system stems from the fact that the competing bandits choose their actions in an asynchronous fashion, and the lower ranked agents only get to learn from a set of arms, not \emph{dominated} by the higher ranked agents, which leads to \emph{forced exploration}. With carefully defined complexity parameters, we characterize this \emph{forced exploration} and obtain sub-linear (logarithmic) regret of \texttt{DNCB}. Furthermore, we validate our theoretical findings via experiments. △ Less

Submitted 31 May, 2022; originally announced June 2022.

Minimax Optimal Online Imitation Learning via Replay Estimation

Authors: Gokul Swamy, Nived Rajaraman, Matthew Peng, Sanjiban Choudhury, J. Andrew Bagnell, Zhiwei Steven Wu, Jiantao Jiao, Kannan Ramchandran

Abstract: Online imitation learning is the problem of how best to mimic expert demonstrations, given access to the environment or an accurate simulator. Prior work has shown that in the infinite sample regime, exact moment matching achieves value equivalence to the expert policy. However, in the finite sample regime, even if one has no optimization error, empirical variance can lead to a performance gap tha… ▽ More Online imitation learning is the problem of how best to mimic expert demonstrations, given access to the environment or an accurate simulator. Prior work has shown that in the infinite sample regime, exact moment matching achieves value equivalence to the expert policy. However, in the finite sample regime, even if one has no optimization error, empirical variance can lead to a performance gap that scales with $H^2 / N$ for behavioral cloning and $H / \sqrt{N}$ for online moment matching, where $H$ is the horizon and $N$ is the size of the expert dataset. We introduce the technique of replay estimation to reduce this empirical variance: by repeatedly executing cached expert actions in a stochastic simulator, we compute a smoother expert visitation distribution estimate to match. In the presence of general function approximation, we prove a meta theorem reducing the performance gap of our approach to the parameter estimation error for offline classification (i.e. learning the expert policy). In the tabular setting or with linear function approximation, our meta theorem shows that the performance gap incurred by our approach achieves the optimal $\widetilde{O} \left( \min({H^{3/2}} / {N}, {H} / {\sqrt{N}} \right)$ dependency, under significantly weaker assumptions compared to prior work. We implement multiple instantiations of our approach on several continuous control tasks and find that we are able to significantly improve policy performance across a variety of dataset sizes. △ Less

Submitted 14 January, 2023; v1 submitted 30 May, 2022; originally announced May 2022.

Evaluating natural language processing models with generalization metrics that do not need access to any training or testing data

Authors: Yaoqing Yang, Ryan Theisen, Liam Hodgkinson, Joseph E. Gonzalez, Kannan Ramchandran, Charles H. Martin, Michael W. Mahoney

Abstract: Selecting suitable architecture parameters and training hyperparameters is essential for enhancing machine learning (ML) model performance. Several recent empirical studies conduct large-scale correlational analysis on neural networks (NNs) to search for effective \emph{generalization metrics} that can guide this type of model selection. Effective metrics are typically expected to correlate strong… ▽ More Selecting suitable architecture parameters and training hyperparameters is essential for enhancing machine learning (ML) model performance. Several recent empirical studies conduct large-scale correlational analysis on neural networks (NNs) to search for effective \emph{generalization metrics} that can guide this type of model selection. Effective metrics are typically expected to correlate strongly with test performance. In this paper, we expand on prior analyses by examining generalization-metric-based model selection with the following objectives: (i) focusing on natural language processing (NLP) tasks, as prior work primarily concentrates on computer vision (CV) tasks; (ii) considering metrics that directly predict \emph{test error} instead of the \emph{generalization gap}; (iii) exploring metrics that do not need access to data to compute. From these objectives, we are able to provide the first model selection results on large pretrained Transformers from Huggingface using generalization metrics. Our analyses consider (I) hundreds of Transformers trained in different settings, in which we systematically vary the amount of data, the model size and the optimization hyperparameters, (II) a total of 51 pretrained Transformers from eight families of Huggingface NLP models, including GPT2, BERT, etc., and (III) a total of 28 existing and novel generalization metrics. Despite their niche status, we find that metrics derived from the heavy-tail (HT) perspective are particularly useful in NLP tasks, exhibiting stronger correlations than other, more popular metrics. To further examine these metrics, we extend prior formulations relying on power law (PL) spectral distributions to exponential (EXP) and exponentially-truncated power law (E-TPL) families. △ Less

Submitted 4 June, 2023; v1 submitted 6 February, 2022; originally announced February 2022.

Journal ref: Proceedings of the 29th ACM SIGKDD international conference on knowledge discovery and data mining (2023)

Taxonomizing local versus global structure in neural network loss landscapes

Authors: Yaoqing Yang, Liam Hodgkinson, Ryan Theisen, Joe Zou, Joseph E. Gonzalez, Kannan Ramchandran, Michael W. Mahoney

Abstract: Viewing neural network models in terms of their loss landscapes has a long history in the statistical mechanics approach to learning, and in recent years it has received attention within machine learning proper. Among other things, local metrics (such as the smoothness of the loss landscape) have been shown to correlate with global properties of the model (such as good generalization performance).… ▽ More Viewing neural network models in terms of their loss landscapes has a long history in the statistical mechanics approach to learning, and in recent years it has received attention within machine learning proper. Among other things, local metrics (such as the smoothness of the loss landscape) have been shown to correlate with global properties of the model (such as good generalization performance). Here, we perform a detailed empirical analysis of the loss landscape structure of thousands of neural network models, systematically varying learning tasks, model architectures, and/or quantity/quality of data. By considering a range of metrics that attempt to capture different aspects of the loss landscape, we demonstrate that the best test accuracy is obtained when: the loss landscape is globally well-connected; ensembles of trained models are more similar to each other; and models converge to locally smooth regions. We also show that globally poorly-connected landscapes can arise when models are small or when they are trained to lower quality data; and that, if the loss landscape is globally poorly-connected, then training to zero loss can actually lead to worse test accuracy. Our detailed empirical results shed light on phases of learning (and consequent double descent behavior), fundamental versus incidental determinants of good generalization, the role of load-like and temperature-like parameters in the learning process, different influences on the loss landscape from model and data, and the relationships between local and global metrics, all topics of recent interest. △ Less

Submitted 12 December, 2021; v1 submitted 23 July, 2021; originally announced July 2021.

Journal ref: Thirty-fifth Annual Conference on Neural Information Processing Systems, 2021

Model Selection for Generic Reinforcement Learning

Authors: Avishek Ghosh, Sayak Ray Chowdhury, Kannan Ramchandran

Abstract: We address the problem of model selection for the finite horizon episodic Reinforcement Learning (RL) problem where the transition kernel $P^*$ belongs to a family of models $\mathcal{P}^*$ with finite metric entropy. In the model selection framework, instead of $\mathcal{P}^*$, we are given $M$ nested families of transition kernels $\cP_1 \subset \cP_2 \subset \ldots \subset \cP_M$. We propose an… ▽ More We address the problem of model selection for the finite horizon episodic Reinforcement Learning (RL) problem where the transition kernel $P^*$ belongs to a family of models $\mathcal{P}^*$ with finite metric entropy. In the model selection framework, instead of $\mathcal{P}^*$, we are given $M$ nested families of transition kernels $\cP_1 \subset \cP_2 \subset \ldots \subset \cP_M$. We propose and analyze a novel algorithm, namely \emph{Adaptive Reinforcement Learning (General)} (\texttt{ARL-GEN}) that adapts to the smallest such family where the true transition kernel $P^*$ lies. \texttt{ARL-GEN} uses the Upper Confidence Reinforcement Learning (\texttt{UCRL}) algorithm with value targeted regression as a blackbox and puts a model selection module at the beginning of each epoch. Under a mild separability assumption on the model classes, we show that \texttt{ARL-GEN} obtains a regret of $\Tilde{\mathcal{O}}(d_{\mathcal{E}}^*H^2+\sqrt{d_{\mathcal{E}}^* \mathbb{M}^* H^2 T})$, with high probability, where $H$ is the horizon length, $T$ is the total number of steps, $d_{\mathcal{E}}^*$ is the Eluder dimension and $\mathbb{M}^*$ is the metric entropy corresponding to $\mathcal{P}^*$. Note that this regret scaling matches that of an oracle that knows $\mathcal{P}^*$ in advance. We show that the cost of model selection for \texttt{ARL-GEN} is an additive term in the regret having a weak dependence on $T$. Subsequently, we remove the separability assumption and consider the setup of linear mixture MDPs, where the transition kernel $P^*$ has a linear function approximation. With this low rank structure, we propose novel adaptive algorithms for model selection, and obtain (order-wise) regret identical to that of an oracle with knowledge of the true model class. △ Less

Submitted 9 December, 2021; v1 submitted 13 July, 2021; originally announced July 2021.

Model Selection for Generic Contextual Bandits

Authors: Avishek Ghosh, Abishek Sankararaman, Kannan Ramchandran

Abstract: We consider the problem of model selection for the general stochastic contextual bandits under the realizability assumption. We propose a successive refinement based algorithm called Adaptive Contextual Bandit ({\ttfamily ACB}), that works in phases and successively eliminates model classes that are too simple to fit the given instance. We prove that this algorithm is adaptive, i.e., the regret ra… ▽ More We consider the problem of model selection for the general stochastic contextual bandits under the realizability assumption. We propose a successive refinement based algorithm called Adaptive Contextual Bandit ({\ttfamily ACB}), that works in phases and successively eliminates model classes that are too simple to fit the given instance. We prove that this algorithm is adaptive, i.e., the regret rate order-wise matches that of any provable contextual bandit algorithm (ex. \cite{falcon}), that needs the knowledge of the true model class. The price of not knowing the correct model class turns out to be only an additive term contributing to the second order term in the regret bound. This cost possess the intuitive property that it becomes smaller as the model class becomes easier to identify, and vice-versa. We also show that a much simpler explore-then-commit (ETC) style algorithm also obtains similar regret bound, despite not knowing the true model class. However, the cost of model selection is higher in ETC as opposed to in {\ttfamily ACB}, as expected. Furthermore, for the special case of linear contextual bandits, we propose specialized algorithms that obtain sharper guarantees compared to the generic setup. △ Less

Submitted 20 July, 2023; v1 submitted 7 July, 2021; originally announced July 2021.

Comments: Accepted at IEEE Transactions on Information Theory. arXiv admin note: text overlap with arXiv:2006.02612

Adaptive Clustering and Personalization in Multi-Agent Stochastic Linear Bandits

Authors: Avishek Ghosh, Abishek Sankararaman, Kannan Ramchandran

Abstract: We consider the problem of minimizing regret in an $N$ agent heterogeneous stochastic linear bandits framework, where the agents (users) are similar but not all identical. We model user heterogeneity using two popularly used ideas in practice; (i) A clustering framework where users are partitioned into groups with users in the same group being identical to each other, but different across groups,… ▽ More We consider the problem of minimizing regret in an $N$ agent heterogeneous stochastic linear bandits framework, where the agents (users) are similar but not all identical. We model user heterogeneity using two popularly used ideas in practice; (i) A clustering framework where users are partitioned into groups with users in the same group being identical to each other, but different across groups, and (ii) a personalization framework where no two users are necessarily identical, but a user's parameters are close to that of the population average. In the clustered users' setup, we propose a novel algorithm, based on successive refinement of cluster identities and regret minimization. We show that, for any agent, the regret scales as $\mathcal{O}(\sqrt{T/N})$, if the agent is in a `well separated' cluster, or scales as $\mathcal{O}(T^{\frac{1}{2} + \varepsilon}/(N)^{\frac{1}{2} -\varepsilon})$ if its cluster is not well separated, where $\varepsilon$ is positive and arbitrarily close to $0$. Our algorithm is adaptive to the cluster separation, and is parameter free -- it does not need to know the number of clusters, separation and cluster size, yet the regret guarantee adapts to the inherent complexity. In the personalization framework, we introduce a natural algorithm where, the personal bandit instances are initialized with the estimates of the global average model. We show that, an agent $i$ whose parameter deviates from the population average by $ε_i$, attains a regret scaling of $\widetilde{O}(ε_i\sqrt{T})$. This demonstrates that if the user representations are close (small $ε_i)$, the resulting regret is low, and vice-versa. The results are empirically validated and we observe superior performance of our adaptive algorithms over non-adaptive baselines. △ Less

Submitted 2 February, 2022; v1 submitted 14 June, 2021; originally announced June 2021.

Comments: 25 pages, 8 figures

LocalNewton: Reducing Communication Bottleneck for Distributed Learning

Authors: Vipul Gupta, Avishek Ghosh, Michal Derezinski, Rajiv Khanna, Kannan Ramchandran, Michael Mahoney

Abstract: To address the communication bottleneck problem in distributed optimization within a master-worker framework, we propose LocalNewton, a distributed second-order algorithm with local averaging. In LocalNewton, the worker machines update their model in every iteration by finding a suitable second-order descent direction using only the data and model stored in their own local memory. We let the worke… ▽ More To address the communication bottleneck problem in distributed optimization within a master-worker framework, we propose LocalNewton, a distributed second-order algorithm with local averaging. In LocalNewton, the worker machines update their model in every iteration by finding a suitable second-order descent direction using only the data and model stored in their own local memory. We let the workers run multiple such iterations locally and communicate the models to the master node only once every few (say L) iterations. LocalNewton is highly practical since it requires only one hyperparameter, the number L of local iterations. We use novel matrix concentration-based techniques to obtain theoretical guarantees for LocalNewton, and we validate them with detailed empirical evaluation. To enhance practicability, we devise an adaptive scheme to choose L, and we show that this reduces the number of local iterations in worker machines between two model synchronizations as the training proceeds, successively refining the model quality at the master. Via extensive experiments using several real-world datasets with AWS Lambda workers and an AWS EC2 master, we show that LocalNewton requires fewer than 60% of the communication rounds (between master and workers) and less than 40% of the end-to-end running time, compared to state-of-the-art algorithms, to reach the same training~loss. △ Less

Submitted 15 May, 2021; originally announced May 2021.

Comments: To be published in Uncertainty in Artificial Intelligence (UAI) 2021

Esca** Saddle Points in Distributed Newton's Method with Communication Efficiency and Byzantine Resilience

Authors: Avishek Ghosh, Raj Kumar Maity, Arya Mazumdar, Kannan Ramchandran

Abstract: The problem of saddle-point avoidance for non-convex optimization is quite challenging in large scale distributed learning frameworks, such as Federated Learning, especially in the presence of Byzantine workers. The celebrated cubic-regularized Newton method of \cite{nest} is one of the most elegant ways to avoid saddle-points in the standard centralized (non-distributed) setup. In this paper, we… ▽ More The problem of saddle-point avoidance for non-convex optimization is quite challenging in large scale distributed learning frameworks, such as Federated Learning, especially in the presence of Byzantine workers. The celebrated cubic-regularized Newton method of \cite{nest} is one of the most elegant ways to avoid saddle-points in the standard centralized (non-distributed) setup. In this paper, we extend the cubic-regularized Newton method to a distributed framework and simultaneously address several practical challenges like communication bottleneck and Byzantine attacks. Note that the issue of saddle-point avoidance becomes more crucial in the presence of Byzantine machines since rogue machines may create \emph{fake local minima} near the saddle-points of the loss function, also known as the saddle-point attack. Being a second order algorithm, our iteration complexity is much lower than the first order counterparts. Furthermore we use compression (or sparsification) techniques like $δ$-approximate compression for communication efficiency. We obtain theoretical guarantees for our proposed scheme under several settings including approximate (sub-sampled) gradients and Hessians. Moreover, we validate our theoretical findings with experiments using standard datasets and several types of Byzantine attacks, and obtain an improvement of $25\%$ with respect to first order methods in iteration complexity. △ Less

Submitted 25 December, 2021; v1 submitted 16 March, 2021; originally announced March 2021.

Provably Breaking the Quadratic Error Compounding Barrier in Imitation Learning, Optimally

Authors: Nived Rajaraman, Yanjun Han, Lin F. Yang, Kannan Ramchandran, Jiantao Jiao

Abstract: We study the statistical limits of Imitation Learning (IL) in episodic Markov Decision Processes (MDPs) with a state space $\mathcal{S}$. We focus on the known-transition setting where the learner is provided a dataset of $N$ length-$H$ trajectories from a deterministic expert policy and knows the MDP transition. We establish an upper bound $O(|\mathcal{S}|H^{3/2}/N)$ for the suboptimality using t… ▽ More We study the statistical limits of Imitation Learning (IL) in episodic Markov Decision Processes (MDPs) with a state space $\mathcal{S}$. We focus on the known-transition setting where the learner is provided a dataset of $N$ length-$H$ trajectories from a deterministic expert policy and knows the MDP transition. We establish an upper bound $O(|\mathcal{S}|H^{3/2}/N)$ for the suboptimality using the Mimic-MD algorithm in Rajaraman et al (2020) which we prove to be computationally efficient. In contrast, we show the minimax suboptimality grows as $Ω( H^{3/2}/N)$ when $|\mathcal{S}|\geq 3$ while the unknown-transition setting suffers from a larger sharp rate $Θ(|\mathcal{S}|H^2/N)$ (Rajaraman et al (2020)). The lower bound is established by proving a two-way reduction between IL and the value estimation problem of the unknown expert policy under any given reward function, as well as building connections with linear functional estimation with subsampled observations. We further show that under the additional assumption that the expert is optimal for the true reward function, there exists an efficient algorithm, which we term as Mimic-Mixture, that provably achieves suboptimality $O(1/N)$ for arbitrary 3-state MDPs with rewards only at the terminal layer. In contrast, no algorithm can achieve suboptimality $O(\sqrt{H}/N)$ with high probability if the expert is not constrained to be optimal. Our work formally establishes the benefit of the expert optimal assumption in the known transition setting, while Rajaraman et al (2020) showed it does not help when transitions are unknown. △ Less

Submitted 25 February, 2021; originally announced February 2021.

Comments: 30 pages, 2 figures

BEAR: Sketching BFGS Algorithm for Ultra-High Dimensional Feature Selection in Sublinear Memory

Authors: Amirali Aghazadeh, Vipul Gupta, Alex DeWeese, O. Ozan Koyluoglu, Kannan Ramchandran

Abstract: We consider feature selection for applications in machine learning where the dimensionality of the data is so large that it exceeds the working memory of the (local) computing machine. Unfortunately, current large-scale sketching algorithms show poor memory-accuracy trade-off due to the irreversible collision and accumulation of the stochastic gradient noise in the sketched domain. Here, we develo… ▽ More We consider feature selection for applications in machine learning where the dimensionality of the data is so large that it exceeds the working memory of the (local) computing machine. Unfortunately, current large-scale sketching algorithms show poor memory-accuracy trade-off due to the irreversible collision and accumulation of the stochastic gradient noise in the sketched domain. Here, we develop a second-order ultra-high dimensional feature selection algorithm, called BEAR, which avoids the extra collisions by storing the second-order gradients in the celebrated Broyden-Fletcher-Goldfarb-Shannon (BFGS) algorithm in Count Sketch, a sublinear memory data structure from the streaming literature. Experiments on real-world data sets demonstrate that BEAR requires up to three orders of magnitude less memory space to achieve the same classification accuracy compared to the first-order sketching algorithms. Theoretical analysis proves convergence of BEAR with rate O(1/t) in t iterations of the sketched algorithm. Our algorithm reveals an unexplored advantage of second-order optimization for memory-constrained sketching of models trained on ultra-high dimensional data sets. △ Less

Submitted 26 May, 2021; v1 submitted 26 October, 2020; originally announced October 2020.

Training Recommender Systems at Scale: Communication-Efficient Model and Data Parallelism

Authors: Vipul Gupta, Dhruv Choudhary, ** Tak Peter Tang, Xiaohan Wei, Xing Wang, Yuzhen Huang, Arun Kejariwal, Kannan Ramchandran, Michael W. Mahoney

Abstract: In this paper, we consider hybrid parallelism -- a paradigm that employs both Data Parallelism (DP) and Model Parallelism (MP) -- to scale distributed training of large recommendation models. We propose a compression framework called Dynamic Communication Thresholding (DCT) for communication-efficient hybrid training. DCT filters the entities to be communicated across the network through a simple… ▽ More In this paper, we consider hybrid parallelism -- a paradigm that employs both Data Parallelism (DP) and Model Parallelism (MP) -- to scale distributed training of large recommendation models. We propose a compression framework called Dynamic Communication Thresholding (DCT) for communication-efficient hybrid training. DCT filters the entities to be communicated across the network through a simple hard-thresholding function, allowing only the most relevant information to pass through. For communication efficient DP, DCT compresses the parameter gradients sent to the parameter server during model synchronization. The threshold is updated only once every few thousand iterations to reduce the computational overhead of compression. For communication efficient MP, DCT incorporates a novel technique to compress the activations and gradients sent across the network during the forward and backward propagation, respectively. This is done by identifying and updating only the most relevant neurons of the neural network for each training sample in the data. We evaluate DCT on publicly available natural language processing and recommender models and datasets, as well as recommendation systems used in production at Facebook. DCT reduces communication by at least $100\times$ and $20\times$ during DP and MP, respectively. The algorithm has been deployed in production, and it improves end-to-end training time for a state-of-the-art industrial recommender model by 37\%, without any loss in performance. △ Less

Submitted 21 May, 2021; v1 submitted 17 October, 2020; originally announced October 2020.

Comments: 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD 2021)

CoVer: Collaborative Light-Node-Only Verification and Data Availability for Blockchains

Authors: Steven Cao, Swanand Kadhe, Kannan Ramchandran

Abstract: Validating a blockchain incurs heavy computation, communication, and storage costs. As a result, clients with limited resources, called light nodes, cannot verify transactions independently and must trust full nodes, making them vulnerable to security attacks. Motivated by this problem, we ask a fundamental question: can light nodes securely validate without any full nodes? We answer affirmatively… ▽ More Validating a blockchain incurs heavy computation, communication, and storage costs. As a result, clients with limited resources, called light nodes, cannot verify transactions independently and must trust full nodes, making them vulnerable to security attacks. Motivated by this problem, we ask a fundamental question: can light nodes securely validate without any full nodes? We answer affirmatively by proposing CoVer, a decentralized protocol that allows a group of light nodes to collaboratively verify blocks even under a dishonest majority, achieving the same level of security for block validation as full nodes while only requiring a fraction of the work. In particular, work per node scales down proportionally with the number of participants (up to a log factor), resulting in computation, communication, and storage requirements that are sublinear in block size. Our main contributions are light-node-only protocols for fraud proofs and data availability. △ Less

Submitted 1 October, 2020; originally announced October 2020.

Comments: IEEE Blockchain 2020

FastSecAgg: Scalable Secure Aggregation for Privacy-Preserving Federated Learning

Authors: Swanand Kadhe, Nived Rajaraman, O. Ozan Koyluoglu, Kannan Ramchandran

Abstract: Recent attacks on federated learning demonstrate that kee** the training data on clients' devices does not provide sufficient privacy, as the model parameters shared by clients can leak information about their training data. A 'secure aggregation' protocol enables the server to aggregate clients' models in a privacy-preserving manner. However, existing secure aggregation protocols incur high com… ▽ More Recent attacks on federated learning demonstrate that kee** the training data on clients' devices does not provide sufficient privacy, as the model parameters shared by clients can leak information about their training data. A 'secure aggregation' protocol enables the server to aggregate clients' models in a privacy-preserving manner. However, existing secure aggregation protocols incur high computation/communication costs, especially when the number of model parameters is larger than the number of clients participating in an iteration -- a typical scenario in federated learning. In this paper, we propose a secure aggregation protocol, FastSecAgg, that is efficient in terms of computation and communication, and robust to client dropouts. The main building block of FastSecAgg is a novel multi-secret sharing scheme, FastShare, based on the Fast Fourier Transform (FFT), which may be of independent interest. FastShare is information-theoretically secure, and achieves a trade-off between the number of secrets, privacy threshold, and dropout tolerance. Riding on the capabilities of FastShare, we prove that FastSecAgg is (i) secure against the server colluding with 'any' subset of some constant fraction (e.g. $\sim10\%$) of the clients in the honest-but-curious setting; and (ii) tolerates dropouts of a 'random' subset of some constant fraction (e.g. $\sim10\%$) of the clients. FastSecAgg achieves significantly smaller computation cost than existing schemes while achieving the same (orderwise) communication cost. In addition, it guarantees security against adaptive adversaries, which can perform client corruptions dynamically during the execution of the protocol. △ Less

Submitted 23 September, 2020; originally announced September 2020.

Comments: Shorter version accepted in ICML Workshop on Federated Learning, July 2020, and CCS Workshop on Privacy-Preserving Machine Learning in Practice, November 2020

Utility-based Resource Allocation and Pricing for Serverless Computing

Authors: Vipul Gupta, Soham Phade, Thomas Courtade, Kannan Ramchandran

Abstract: Serverless computing platforms currently rely on basic pricing schemes that are static and do not reflect customer feedback. This leads to significant inefficiencies from a total utility perspective. As one of the fastest-growing cloud services, serverless computing provides an opportunity to better serve both users and providers through the incorporation of market-based strategies for pricing and… ▽ More Serverless computing platforms currently rely on basic pricing schemes that are static and do not reflect customer feedback. This leads to significant inefficiencies from a total utility perspective. As one of the fastest-growing cloud services, serverless computing provides an opportunity to better serve both users and providers through the incorporation of market-based strategies for pricing and resource allocation. With the help of utility functions to model the delay-sensitivity of customers, we propose a novel scheduler to allocate resources for serverless computing. The resulting resource allocation scheme is optimal in the sense that it maximizes the aggregate utility of all users across the system, thus maximizing social welfare. Our approach gives rise to a natural dynamic pricing scheme that is obtained by solving an optimization problem in its dual form. We further develop feedback mechanisms that allow the cloud provider to converge to optimal resource allocation, even when the users' utilities are private and unknown to the service provider. Simulations show that our approach can track market demand and achieve significantly higher social welfare (or, equivalently, cost savings for customers) compared to existing schemes. △ Less

Submitted 24 January, 2022; v1 submitted 18 August, 2020; originally announced August 2020.

Comments: 31 pages, 10 figures

Boundary thickness and robustness in learning models

Authors: Yaoqing Yang, Rajiv Khanna, Yaodong Yu, Amir Gholami, Kurt Keutzer, Joseph E. Gonzalez, Kannan Ramchandran, Michael W. Mahoney

Abstract: Robustness of machine learning models to various adversarial and non-adversarial corruptions continues to be of interest. In this paper, we introduce the notion of the boundary thickness of a classifier, and we describe its connection with and usefulness for model robustness. Thick decision boundaries lead to improved performance, while thin decision boundaries lead to overfitting (e.g., measured… ▽ More Robustness of machine learning models to various adversarial and non-adversarial corruptions continues to be of interest. In this paper, we introduce the notion of the boundary thickness of a classifier, and we describe its connection with and usefulness for model robustness. Thick decision boundaries lead to improved performance, while thin decision boundaries lead to overfitting (e.g., measured by the robust generalization gap between training and testing) and lower robustness. We show that a thicker boundary helps improve robustness against adversarial examples (e.g., improving the robust test accuracy of adversarial training) as well as so-called out-of-distribution (OOD) transforms, and we show that many commonly-used regularization and data augmentation procedures can increase boundary thickness. On the theoretical side, we establish that maximizing boundary thickness during training is akin to the so-called mixup training. Using these observations, we show that noise-augmentation on mixup training further increases boundary thickness, thereby combating vulnerability to various forms of adversarial attacks and OOD transforms. We can also show that the performance improvement in several lines of recent work happens in conjunction with a thicker boundary. △ Less

Submitted 12 January, 2021; v1 submitted 9 July, 2020; originally announced July 2020.

Journal ref: NeurIPS 2020

An Efficient Framework for Clustered Federated Learning

Authors: Avishek Ghosh, Jichan Chung, Dong Yin, Kannan Ramchandran

Abstract: We address the problem of federated learning (FL) where users are distributed and partitioned into clusters. This setup captures settings where different groups of users have their own objectives (learning tasks) but by aggregating their data with others in the same cluster (same learning task), they can leverage the strength in numbers in order to perform more efficient federated learning. For th… ▽ More We address the problem of federated learning (FL) where users are distributed and partitioned into clusters. This setup captures settings where different groups of users have their own objectives (learning tasks) but by aggregating their data with others in the same cluster (same learning task), they can leverage the strength in numbers in order to perform more efficient federated learning. For this new framework of clustered federated learning, we propose the Iterative Federated Clustering Algorithm (IFCA), which alternately estimates the cluster identities of the users and optimizes model parameters for the user clusters via gradient descent. We analyze the convergence rate of this algorithm first in a linear model with squared loss and then for generic strongly convex and smooth loss functions. We show that in both settings, with good initialization, IFCA is guaranteed to converge, and discuss the optimality of the statistical error rate. In particular, for the linear model with two clusters, we can guarantee that our algorithm converges as long as the initialization is slightly better than random. When the clustering structure is ambiguous, we propose to train the models by combining IFCA with the weight sharing technique in multi-task learning. In the experiments, we show that our algorithm can succeed even if we relax the requirements on initialization with random initialization and multiple restarts. We also present experimental results showing that our algorithm is efficient in non-convex problems such as neural networks. We demonstrate the benefits of IFCA over the baselines on several clustered FL benchmarks. △ Less

Submitted 8 June, 2021; v1 submitted 7 June, 2020; originally announced June 2020.

Comments: Preliminary results appeared at NeurIPS 2020

Problem-Complexity Adaptive Model Selection for Stochastic Linear Bandits

Authors: Avishek Ghosh, Abishek Sankararaman, Kannan Ramchandran

Abstract: We consider the problem of model selection for two popular stochastic linear bandit settings, and propose algorithms that adapts to the unknown problem complexity. In the first setting, we consider the $K$ armed mixture bandits, where the mean reward of arm $i \in [K]$, is $μ_i+ \langle α_{i,t},θ^* \rangle $, with $α_{i,t} \in \mathbb{R}^d$ being the known context vector and $μ_i \in [-1,1]$ and… ▽ More We consider the problem of model selection for two popular stochastic linear bandit settings, and propose algorithms that adapts to the unknown problem complexity. In the first setting, we consider the $K$ armed mixture bandits, where the mean reward of arm $i \in [K]$, is $μ_i+ \langle α_{i,t},θ^* \rangle $, with $α_{i,t} \in \mathbb{R}^d$ being the known context vector and $μ_i \in [-1,1]$ and $θ^*$ are unknown parameters. We define $\|θ^*\|$ as the problem complexity and consider a sequence of nested hypothesis classes, each positing a different upper bound on $\|θ^*\|$. Exploiting this, we propose Adaptive Linear Bandit (ALB), a novel phase based algorithm that adapts to the true problem complexity, $\|θ^*\|$. We show that ALB achieves regret scaling of $O(\|θ^*\|\sqrt{T})$, where $\|θ^*\|$ is apriori unknown. As a corollary, when $θ^*=0$, ALB recovers the minimax regret for the simple bandit algorithm without such knowledge of $θ^*$. ALB is the first algorithm that uses parameter norm as model section criteria for linear bandits. Prior state of art algorithms \cite{osom} achieve a regret of $O(L\sqrt{T})$, where $L$ is the upper bound on $\|θ^*\|$, fed as an input to the problem. In the second setting, we consider the standard linear bandit problem (with possibly an infinite number of arms) where the sparsity of $θ^*$, denoted by $d^* \leq d$, is unknown to the algorithm. Defining $d^*$ as the problem complexity, we show that ALB achieves $O(d^*\sqrt{T})$ regret, matching that of an oracle who knew the true sparsity level. This methodology is then extended to the case of finitely many arms and similar results are proven. This is the first algorithm that achieves such model selection guarantees. We further verify our results via synthetic and real-data experiments. △ Less

Submitted 15 June, 2020; v1 submitted 3 June, 2020; originally announced June 2020.

Comments: 24 pages, 8 figures

Communication-Efficient Gradient Coding for Straggler Mitigation in Distributed Learning

Authors: Swanand Kadhe, O. Ozan Koyluoglu, Kannan Ramchandran

Abstract: Distributed implementations of gradient-based methods, wherein a server distributes gradient computations across worker machines, need to overcome two limitations: delays caused by slow running machines called 'stragglers', and communication overheads. Recently, Ye and Abbe [ICML 2018] proposed a coding-theoretic paradigm to characterize a fundamental trade-off between computation load per worker,… ▽ More Distributed implementations of gradient-based methods, wherein a server distributes gradient computations across worker machines, need to overcome two limitations: delays caused by slow running machines called 'stragglers', and communication overheads. Recently, Ye and Abbe [ICML 2018] proposed a coding-theoretic paradigm to characterize a fundamental trade-off between computation load per worker, communication overhead per worker, and straggler tolerance. However, their proposed coding schemes suffer from heavy decoding complexity and poor numerical stability. In this paper, we develop a communication-efficient gradient coding framework to overcome these drawbacks. Our proposed framework enables using any linear code to design the encoding and decoding functions. When a particular code is used in this framework, its block-length determines the computation load, dimension determines the communication overhead, and minimum distance determines the straggler tolerance. The flexibility of choosing a code allows us to gracefully trade-off the straggler threshold and communication overhead for smaller decoding complexity and higher numerical stability. Further, we show that using a maximum distance separable (MDS) code generated by a random Gaussian matrix in our framework yields a gradient code that is optimal with respect to the trade-off and, in addition, satisfies stronger guarantees on numerical stability as compared to the previously proposed schemes. Finally, we evaluate our proposed framework on Amazon EC2 and demonstrate that it reduces the average iteration time by 16% as compared to prior gradient coding schemes. △ Less

Submitted 14 May, 2020; originally announced May 2020.

Comments: Shorter version accepted in 2020 IEEE International Symposium on Information Theory (ISIT)

Alternating Minimization Converges Super-Linearly for Mixed Linear Regression

Authors: Avishek Ghosh, Kannan Ramchandran

Abstract: We address the problem of solving mixed random linear equations. We have unlabeled observations coming from multiple linear regressions, and each observation corresponds to exactly one of the regression models. The goal is to learn the linear regressors from the observations. Classically, Alternating Minimization (AM) (which is a variant of Expectation Maximization (EM)) is used to solve this prob… ▽ More We address the problem of solving mixed random linear equations. We have unlabeled observations coming from multiple linear regressions, and each observation corresponds to exactly one of the regression models. The goal is to learn the linear regressors from the observations. Classically, Alternating Minimization (AM) (which is a variant of Expectation Maximization (EM)) is used to solve this problem. AM iteratively alternates between the estimation of labels and solving the regression problems with the estimated labels. Empirically, it is observed that, for a large variety of non-convex problems including mixed linear regression, AM converges at a much faster rate compared to gradient based algorithms. However, the existing theory suggests similar rate of convergence for AM and gradient based methods, failing to capture this empirical behavior. In this paper, we close this gap between theory and practice for the special case of a mixture of $2$ linear regressions. We show that, provided initialized properly, AM enjoys a \emph{super-linear} rate of convergence in certain parameter regimes. To the best of our knowledge, this is the first work that theoretically establishes such rate for AM. Hence, if we want to recover the unknown regressors upto an error (in $\ell_2$ norm) of $ε$, AM only takes $\mathcal{O}(\log \log (1/ε))$ iterations. Furthermore, we compare AM with a gradient based heuristic algorithm empirically and show that AM dominates in iteration complexity as well as wall-clock time. △ Less

Submitted 11 August, 2020; v1 submitted 22 April, 2020; originally announced April 2020.

Comments: Accepted for publication at AISTATS, 2020

Serverless Straggler Mitigation using Local Error-Correcting Codes

Authors: Vipul Gupta, Dominic Carrano, Yaoqing Yang, Vaishaal Shankar, Thomas Courtade, Kannan Ramchandran

Abstract: Inexpensive cloud services, such as serverless computing, are often vulnerable to straggling nodes that increase end-to-end latency for distributed computation. We propose and implement simple yet principled approaches for straggler mitigation in serverless systems for matrix multiplication and evaluate them on several common applications from machine learning and high-performance computing. The p… ▽ More Inexpensive cloud services, such as serverless computing, are often vulnerable to straggling nodes that increase end-to-end latency for distributed computation. We propose and implement simple yet principled approaches for straggler mitigation in serverless systems for matrix multiplication and evaluate them on several common applications from machine learning and high-performance computing. The proposed schemes are inspired by error-correcting codes and employ parallel encoding and decoding over the data stored in the cloud using serverless workers. This creates a fully distributed computing framework without using a master node to conduct encoding or decoding, which removes the computation, communication and storage bottleneck at the master. On the theory side, we establish that our proposed scheme is asymptotically optimal in terms of decoding time and provide a lower bound on the number of stragglers it can tolerate with high probability. Through extensive experiments, we show that our scheme outperforms existing schemes such as speculative execution and other coding theoretic methods by at least 25%. △ Less

Submitted 21 January, 2020; originally announced January 2020.

Communication-Efficient and Byzantine-Robust Distributed Learning with Error Feedback

Authors: Avishek Ghosh, Raj Kumar Maity, Swanand Kadhe, Arya Mazumdar, Kannan Ramchandran

Abstract: We develop a communication-efficient distributed learning algorithm that is robust against Byzantine worker machines. We propose and analyze a distributed gradient-descent algorithm that performs a simple thresholding based on gradient norms to mitigate Byzantine failures. We show the (statistical) error-rate of our algorithm matches that of Yin et al.~\cite{dong}, which uses more complicated sche… ▽ More We develop a communication-efficient distributed learning algorithm that is robust against Byzantine worker machines. We propose and analyze a distributed gradient-descent algorithm that performs a simple thresholding based on gradient norms to mitigate Byzantine failures. We show the (statistical) error-rate of our algorithm matches that of Yin et al.~\cite{dong}, which uses more complicated schemes (coordinate-wise median, trimmed mean). Furthermore, for communication efficiency, we consider a generic class of $δ$-approximate compressors from Karimireddi et al.~\cite{errorfeed} that encompasses sign-based compressors and top-$k$ sparsification. Our algorithm uses compressed gradients and gradient norms for aggregation and Byzantine removal respectively. We establish the statistical error rate for non-convex smooth loss functions. We show that, in certain range of the compression factor $δ$, the (order-wise) rate of convergence is not affected by the compression operation. Moreover, we analyze the compressed gradient descent algorithm with error feedback (proposed in \cite{errorfeed}) in a distributed setting and in the presence of Byzantine worker machines. We show that exploiting error feedback improves the statistical error rate. Finally, we experimentally validate our results and show good performance in convergence for convex (least-square regression) and non-convex (neural network training) problems. △ Less

Submitted 14 August, 2021; v1 submitted 21 November, 2019; originally announced November 2019.

SeF: A Secure Fountain Architecture for Slashing Storage Costs in Blockchains

Authors: Swanand Kadhe, Jichan Chung, Kannan Ramchandran

Abstract: Full nodes, which synchronize the entire blockchain history and independently validate all the blocks, form the backbone of any blockchain network by playing a vital role in ensuring security properties. On the other hand, a user running a full node needs to pay a heavy price in terms of storage costs. E.g., the Bitcoin blockchain size has grown over 215GB, in spite of its low throughput. The ledg… ▽ More Full nodes, which synchronize the entire blockchain history and independently validate all the blocks, form the backbone of any blockchain network by playing a vital role in ensuring security properties. On the other hand, a user running a full node needs to pay a heavy price in terms of storage costs. E.g., the Bitcoin blockchain size has grown over 215GB, in spite of its low throughput. The ledger size for a high throughput blockchain Ripple has already reached 9TB, and it is growing at an astonishing rate of 12GB per day! In this paper, we propose an architecture based on 'fountain codes', a class of erasure codes, that enables any full node to 'encode' validated blocks into a small number of 'coded blocks', thereby reducing its storage costs by orders of magnitude. In particular, our proposed "Secure Fountain (SeF)" architecture can achieve a near-optimal trade-off between the storage savings per node and the 'bootstrap cost' in terms of the number of (honest) storage-constrained nodes a new node needs to contact to recover the blockchain. A key technical innovation in SeF codes is to make fountain codes secure against adversarial nodes that can provide maliciously formed coded blocks. Our idea is to use the header-chain as a 'side-information' to check whether a coded block is maliciously formed while it is getting decoded. Further, the 'rateless property' of fountain codes helps in achieving high decentralization and scalability. Our experiments demonstrate that SeF codes tuned to achieve 1000x storage savings enable full nodes to encode the 191GB Bitcoin blockchain into 195MB on average. A new node can recover the blockchain from an arbitrary set of storage-constrained nodes as long as the set contains ~1100 honest nodes on average. Note that for a 1000x storage savings, the fundamental bound on the number of honest nodes to contact is 1000: we need about 10% more in practice. △ Less

Submitted 28 June, 2019; originally announced June 2019.

Max-Affine Regression: Provable, Tractable, and Near-Optimal Statistical Estimation

Authors: Avishek Ghosh, Ashwin Pananjady, Adityanand Guntuboyina, Kannan Ramchandran

Abstract: Max-affine regression refers to a model where the unknown regression function is modeled as a maximum of $k$ unknown affine functions for a fixed $k \geq 1$. This generalizes linear regression and (real) phase retrieval, and is closely related to convex regression. Working within a non-asymptotic framework, we study this problem in the high-dimensional setting assuming that $k$ is a fixed constant… ▽ More Max-affine regression refers to a model where the unknown regression function is modeled as a maximum of $k$ unknown affine functions for a fixed $k \geq 1$. This generalizes linear regression and (real) phase retrieval, and is closely related to convex regression. Working within a non-asymptotic framework, we study this problem in the high-dimensional setting assuming that $k$ is a fixed constant, and focus on estimation of the unknown coefficients of the affine functions underlying the model. We analyze a natural alternating minimization (AM) algorithm for the non-convex least squares objective when the design is random. We show that the AM algorithm, when initialized suitably, converges with high probability and at a geometric rate to a small ball around the optimal coefficients. In order to initialize the algorithm, we propose and analyze a combination of a spectral method and a random search scheme in a low-dimensional space, which may be of independent interest. The final rate that we obtain is near-parametric and minimax optimal (up to a poly-logarithmic factor) as a function of the dimension, sample size, and noise variance. In that sense, our approach should be viewed as a direct and implementable method of enforcing regularization to alleviate the curse of dimensionality in problems of the convex regression type. As a by-product of our analysis, we also obtain guarantees on a classical algorithm for the phase retrieval problem under considerably weaker assumptions on the design distribution than was previously known. Numerical experiments illustrate the sharpness of our bounds in the various problem parameters. △ Less

Submitted 21 June, 2019; originally announced June 2019.

Comments: The first two authors contributed equally to this work and are ordered alphabetically

Robust Federated Learning in a Heterogeneous Environment

Authors: Avishek Ghosh, Justin Hong, Dong Yin, Kannan Ramchandran

Abstract: We study a recently proposed large-scale distributed learning paradigm, namely Federated Learning, where the worker machines are end users' own devices. Statistical and computational challenges arise in Federated Learning particularly in the presence of heterogeneous data distribution (i.e., data points on different devices belong to different distributions signifying different clusters) and Byzan… ▽ More We study a recently proposed large-scale distributed learning paradigm, namely Federated Learning, where the worker machines are end users' own devices. Statistical and computational challenges arise in Federated Learning particularly in the presence of heterogeneous data distribution (i.e., data points on different devices belong to different distributions signifying different clusters) and Byzantine machines (i.e., machines that may behave abnormally, or even exhibit arbitrary and potentially adversarial behavior). To address the aforementioned challenges, first we propose a general statistical model for this problem which takes both the cluster structure of the users and the Byzantine machines into account. Then, leveraging the statistical model, we solve the robust heterogeneous Federated Learning problem \emph{optimally}; in particular our algorithm matches the lower bound on the estimation error in dimension and the number of data points. Furthermore, as a by-product, we prove statistical guarantees for an outlier-robust clustering algorithm, which can be considered as the Lloyd algorithm with robust estimation. Finally, we show via synthetic as well as real data experiments that the estimation error obtained by our proposed algorithm is significantly better than the non-Byzantine-robust algorithms; in particular, we gain at least by 53\% and 33\% for synthetic and real data experiments, respectively, in typical settings. △ Less

Submitted 9 October, 2019; v1 submitted 15 June, 2019; originally announced June 2019.

Comments: Fixing technical issues. Please discard any previous version

Adversarially Trained Autoencoders for Parallel-Data-Free Voice Conversion

Authors: Orhan Ocal, Oguz H. Elibol, Gokce Keskin, Cory Stephenson, Anil Thomas, Kannan Ramchandran

Abstract: We present a method for converting the voices between a set of speakers. Our method is based on training multiple autoencoder paths, where there is a single speaker-independent encoder and multiple speaker-dependent decoders. The autoencoders are trained with an addition of an adversarial loss which is provided by an auxiliary classifier in order to guide the output of the encoder to be speaker in… ▽ More We present a method for converting the voices between a set of speakers. Our method is based on training multiple autoencoder paths, where there is a single speaker-independent encoder and multiple speaker-dependent decoders. The autoencoders are trained with an addition of an adversarial loss which is provided by an auxiliary classifier in order to guide the output of the encoder to be speaker independent. The training of the model is unsupervised in the sense that it does not require collecting the same utterances from the speakers nor does it require time aligning over phonemes. Due to the use of a single encoder, our method can generalize to converting the voice of out-of-training speakers to speakers in the training dataset. We present subjective tests corroborating the performance of our method. △ Less

Submitted 9 May, 2019; originally announced May 2019.

Gradient Coding Based on Block Designs for Mitigating Adversarial Stragglers

Authors: Swanand Kadhe, O. Ozan Koyluoglu, Kannan Ramchandran

Abstract: Distributed implementations of gradient-based methods, wherein a server distributes gradient computations across worker machines, suffer from slow running machines, called 'stragglers'. Gradient coding is a coding-theoretic framework to mitigate stragglers by enabling the server to recover the gradient sum in the presence of stragglers. 'Approximate gradient codes' are variants of gradient codes t… ▽ More Distributed implementations of gradient-based methods, wherein a server distributes gradient computations across worker machines, suffer from slow running machines, called 'stragglers'. Gradient coding is a coding-theoretic framework to mitigate stragglers by enabling the server to recover the gradient sum in the presence of stragglers. 'Approximate gradient codes' are variants of gradient codes that reduce computation and storage overhead per worker by allowing the server to approximately reconstruct the gradient sum. In this work, our goal is to construct approximate gradient codes that are resilient to stragglers selected by a computationally unbounded adversary. Our motivation for constructing codes to mitigate adversarial stragglers stems from the challenge of tackling stragglers in massive-scale elastic and serverless systems, wherein it is difficult to statistically model stragglers. Towards this end, we propose a class of approximate gradient codes based on balanced incomplete block designs (BIBDs). We show that the approximation error for these codes depends only on the number of stragglers, and thus, adversarial straggler selection has no advantage over random selection. In addition, the proposed codes admit computationally efficient decoding at the server. Next, to characterize fundamental limits of adversarial straggling, we consider the notion of 'adversarial threshold' -- the smallest number of workers that an adversary must straggle to inflict certain approximation error. We compute a lower bound on the adversarial threshold, and show that codes based on symmetric BIBDs maximize this lower bound among a wide class of codes, making them excellent candidates for mitigating adversarial stragglers. △ Less

Submitted 30 April, 2019; originally announced April 2019.

Comments: Shorter version accepted in 2019 IEEE International Symposium on Information Theory (ISIT)

OverSketched Newton: Fast Convex Optimization for Serverless Systems

Authors: Vipul Gupta, Swanand Kadhe, Thomas Courtade, Michael W. Mahoney, Kannan Ramchandran

Abstract: Motivated by recent developments in serverless systems for large-scale computation as well as improvements in scalable randomized matrix algorithms, we develop OverSketched Newton, a randomized Hessian-based optimization algorithm to solve large-scale convex optimization problems in serverless systems. OverSketched Newton leverages matrix sketching ideas from Randomized Numerical Linear Algebra to… ▽ More Motivated by recent developments in serverless systems for large-scale computation as well as improvements in scalable randomized matrix algorithms, we develop OverSketched Newton, a randomized Hessian-based optimization algorithm to solve large-scale convex optimization problems in serverless systems. OverSketched Newton leverages matrix sketching ideas from Randomized Numerical Linear Algebra to compute the Hessian approximately. These sketching methods lead to inbuilt resiliency against stragglers that are a characteristic of serverless architectures. Depending on whether the problem is strongly convex or not, we propose different iteration updates using the approximate Hessian. For both cases, we establish convergence guarantees for OverSketched Newton and empirically validate our results by solving large-scale supervised learning problems on real-world datasets. Experiments demonstrate a reduction of ~50% in total running time on AWS Lambda, compared to state-of-the-art distributed optimization schemes. △ Less

Submitted 27 August, 2020; v1 submitted 21 March, 2019; originally announced March 2019.

Comments: 37 pages, 12 figures

Cross-Entropy Loss and Low-Rank Features Have Responsibility for Adversarial Examples

Authors: Kamil Nar, Orhan Ocal, S. Shankar Sastry, Kannan Ramchandran

Abstract: State-of-the-art neural networks are vulnerable to adversarial examples; they can easily misclassify inputs that are imperceptibly different than their training and test data. In this work, we establish that the use of cross-entropy loss function and the low-rank features of the training data have responsibility for the existence of these inputs. Based on this observation, we suggest that addressi… ▽ More State-of-the-art neural networks are vulnerable to adversarial examples; they can easily misclassify inputs that are imperceptibly different than their training and test data. In this work, we establish that the use of cross-entropy loss function and the low-rank features of the training data have responsibility for the existence of these inputs. Based on this observation, we suggest that addressing adversarial examples requires rethinking the use of cross-entropy loss function and looking for an alternative that is more suited for minimization with low-rank features. In this direction, we present a training scheme called differential training, which uses a loss function defined on the differences between the features of points from opposite classes. We show that differential training can ensure a large margin between the decision boundary of the neural network and the points in the training dataset. This larger margin increases the amount of perturbation needed to flip the prediction of the classifier and makes it harder to find an adversarial example with small perturbations. We test differential training on a binary classification task with CIFAR-10 dataset and demonstrate that it radically reduces the ratio of images for which an adversarial example could be found -- not only in the training dataset, but in the test dataset as well. △ Less

Submitted 24 January, 2019; originally announced January 2019.

Greedy Frank-Wolfe Algorithm for Exemplar Selection

Authors: Gary Cheng, Armin Askari, Kannan Ramchandran, Laurent El Ghaoui

Abstract: In this paper, we consider the problem of selecting representatives from a data set for arbitrary supervised/unsupervised learning tasks. We identify a subset $S$ of a data set $A$ such that 1) the size of $S$ is much smaller than $A$ and 2) $S$ efficiently describes the entire data set, in a way formalized via convex optimization. In order to generate $|S| = k$ exemplars, our kernelizable algorit… ▽ More In this paper, we consider the problem of selecting representatives from a data set for arbitrary supervised/unsupervised learning tasks. We identify a subset $S$ of a data set $A$ such that 1) the size of $S$ is much smaller than $A$ and 2) $S$ efficiently describes the entire data set, in a way formalized via convex optimization. In order to generate $|S| = k$ exemplars, our kernelizable algorithm, Frank-Wolfe Sparse Representation (FWSR), only needs to execute $\approx k$ iterations with a per-iteration cost that is quadratic in the size of $A$. This is in contrast to other state of the art methods which need to execute until convergence with each iteration costing an extra factor of $d$ (dimension of the data). Moreover, we also provide a proof of linear convergence for our method. We support our results with empirical experiments; we test our algorithm against current methods in three different experimental setups on four different data sets. FWSR outperforms other exemplar finding methods both in speed and accuracy in almost all scenarios. △ Less

Submitted 22 February, 2020; v1 submitted 6 November, 2018; originally announced November 2018.

OverSketch: Approximate Matrix Multiplication for the Cloud

Authors: Vipul Gupta, Shusen Wang, Thomas Courtade, Kannan Ramchandran

Abstract: We propose OverSketch, an approximate algorithm for distributed matrix multiplication in serverless computing. OverSketch leverages ideas from matrix sketching and high-performance computing to enable cost-efficient multiplication that is resilient to faults and straggling nodes pervasive in low-cost serverless architectures. We establish statistical guarantees on the accuracy of OverSketch and em… ▽ More We propose OverSketch, an approximate algorithm for distributed matrix multiplication in serverless computing. OverSketch leverages ideas from matrix sketching and high-performance computing to enable cost-efficient multiplication that is resilient to faults and straggling nodes pervasive in low-cost serverless architectures. We establish statistical guarantees on the accuracy of OverSketch and empirically validate our results by solving a large-scale linear program using interior-point methods and demonstrate a 34% reduction in compute time on AWS Lambda. △ Less

Submitted 21 February, 2019; v1 submitted 6 November, 2018; originally announced November 2018.

Comments: Published in Proc. IEEE Big Data 2018. Updated version provides details of distributed sketching and highlights other advantages of OverSketch

Journal ref: 2018 IEEE International Conference on Big Data (Big Data), Seattle, WA, USA, 2018, pp. 298-304

Rademacher Complexity for Adversarially Robust Generalization

Authors: Dong Yin, Kannan Ramchandran, Peter Bartlett

Abstract: Many machine learning models are vulnerable to adversarial attacks; for example, adding adversarial perturbations that are imperceptible to humans can often make machine learning models produce wrong predictions with high confidence. Moreover, although we may obtain robust models on the training dataset via adversarial training, in some problems the learned models cannot generalize well to the tes… ▽ More Many machine learning models are vulnerable to adversarial attacks; for example, adding adversarial perturbations that are imperceptible to humans can often make machine learning models produce wrong predictions with high confidence. Moreover, although we may obtain robust models on the training dataset via adversarial training, in some problems the learned models cannot generalize well to the test data. In this paper, we focus on $\ell_\infty$ attacks, and study the adversarially robust generalization problem through the lens of Rademacher complexity. For binary linear classifiers, we prove tight bounds for the adversarial Rademacher complexity, and show that the adversarial Rademacher complexity is never smaller than its natural counterpart, and it has an unavoidable dimension dependence, unless the weight vector has bounded $\ell_1$ norm. The results also extend to multi-class linear classifiers. For (nonlinear) neural networks, we show that the dimension dependence in the adversarial Rademacher complexity also exists. We further consider a surrogate adversarial loss for one-hidden layer ReLU network and prove margin bounds for this setting. Our results indicate that having $\ell_1$ norm constraints on the weight matrices might be a potential way to improve generalization in the adversarial setting. We demonstrate experimental results that validate our theoretical findings. △ Less

Submitted 29 July, 2020; v1 submitted 28 October, 2018; originally announced October 2018.

Comments: ICML 2019

Online Scoring with Delayed Information: A Convex Optimization Viewpoint

Authors: Avishek Ghosh, Kannan Ramchandran

Abstract: We consider a system where agents enter in an online fashion and are evaluated based on their attributes or context vectors. There can be practical situations where this context is partially observed, and the unobserved part comes after some delay. We assume that an agent, once left, cannot re-enter the system. Therefore, the job of the system is to provide an estimated score for the agent based o… ▽ More We consider a system where agents enter in an online fashion and are evaluated based on their attributes or context vectors. There can be practical situations where this context is partially observed, and the unobserved part comes after some delay. We assume that an agent, once left, cannot re-enter the system. Therefore, the job of the system is to provide an estimated score for the agent based on her instantaneous score and possibly some inference of the instantaneous score over the delayed score. In this paper, we estimate the delayed context via an online convex game between the agent and the system. We argue that the error in the score estimate accumulated over $T$ iterations is small if the regret of the online convex game is small. Further, we leverage side information about the delayed context in the form of a correlation function with the known context. We consider the settings where the delay is fixed or arbitrarily chosen by an adversary. Furthermore, we extend the formulation to the setting where the contexts are drawn from some Banach space. Overall, we show that the average penalty for not knowing the delayed context while making a decision scales with $\mathcal{O}(\frac{1}{\sqrt{T}})$, where this can be improved to $\mathcal{O}(\frac{\log T}{T})$ under special setting. △ Less

Submitted 9 July, 2018; originally announced July 2018.

Comments: 8 pages, 4 figures

Faster Data-access in Large-scale Systems: Network-scale Latency Analysis under General Service-time Distributions

Authors: Avishek Ghosh, Kannan Ramchandran

Abstract: In cloud storage systems with a large number of servers, files are typically not stored in single servers. Instead, they are split, replicated (to ensure reliability in case of server malfunction) and stored in different servers. We analyze the mean latency of such a split-and-replicate cloud storage system under general sub-exponential service time. We present a novel scheduling scheme that utili… ▽ More In cloud storage systems with a large number of servers, files are typically not stored in single servers. Instead, they are split, replicated (to ensure reliability in case of server malfunction) and stored in different servers. We analyze the mean latency of such a split-and-replicate cloud storage system under general sub-exponential service time. We present a novel scheduling scheme that utilizes the load-balancing policy of the \textit{power of $d$ $(\geq 2)$} choices. An alternative to split-and-replicate is to use erasure-codes, and recently, it has been observed that they can reduce latency in data access (see \cite{longbo_delay} for details). We argue that under high redundancy (integer redundancy factor strictly greater than or equal to 2) regime, the mean latency of a coded system is upper bounded by that of a split-and-replicate system (with same replication factor) and the gap between these two is small. We validate this claim numerically under different service distributions such as exponential, shift plus exponential and the heavy-tailed Weibull distribution and compare the mean latency to that of an unsplit-replicated system. We observe that the coded system outperforms the unsplit-replication system by at least $20\%$. Furthermore, we consider the mean latency for an erasure coded system with low redundancy (fractional redundancy factor between 1 and 2), a scenario which is more pragmatic, given the storage constraints (\cite{rashmi_thesis}). However under this regime, we restrict ourselves to the special case of exponential service time distribution and use the randomized load balancing policy namely \textit{batch-sampling}. We obtain an upper bound on mean delay that depends on the order statistics of the queue lengths, which, we further smooth out via a discrete to continuous approximation. △ Less

Submitted 6 July, 2018; originally announced July 2018.

Comments: 12 pages, 7 figures

Matching Observations to Distributions: Efficient Estimation via Sparsified Hungarian Algorithm

Authors: Sinho Chewi, Forest Yang, Avishek Ghosh, Abhay Parekh, Kannan Ramchandran

Abstract: Suppose we are given observations, where each observation is drawn independently from one of $k$ known distributions. The goal is to match each observation to the distribution from which it was drawn. We observe that the maximum likelihood estimator (MLE) for this problem can be computed using weighted bipartite matching, even when $n$, the number of observations per distribution, exceeds one. Thi… ▽ More Suppose we are given observations, where each observation is drawn independently from one of $k$ known distributions. The goal is to match each observation to the distribution from which it was drawn. We observe that the maximum likelihood estimator (MLE) for this problem can be computed using weighted bipartite matching, even when $n$, the number of observations per distribution, exceeds one. This is achieved by instantiating $n$ duplicates of each distribution node. However, in the regime where the number of observations per distribution is much larger than the number of distributions, the Hungarian matching algorithm for computing the weighted bipartite matching requires $\mathcal O(n^3)$ time. We introduce a novel randomized matching algorithm that reduces the runtime to $\tilde{\mathcal O}(n^2)$ by sparsifying the original graph, returning the exact MLE with high probability. Next, we give statistical justification for using the MLE by bounding the excess risk of the MLE, where the loss is defined as the negative log-likelihood. We test these bounds for the case of isotropic Gaussians with equal covariances and whose means are separated by a distance $η$, and find (1) that $\gg \log k$ separation suffices to drive the proportion of mismatches of the MLE to 0, and (2) that the expected fraction of mismatched observations goes to zero at rate $\mathcal O({(\log k)}^2/η^2)$. △ Less

Submitted 29 September, 2019; v1 submitted 18 June, 2018; originally announced June 2018.

Comments: 8 pages, 1 figure; to appear in the 57th Annual Allerton Conference on Communication, Control, and Computing

MSC Class: 68W20

Customized Local Differential Privacy for Multi-Agent Distributed Optimization

Authors: Roel Dobbe, Ye Pu, **gge Zhu, Kannan Ramchandran, Claire Tomlin

Abstract: Real-time data-driven optimization and control problems over networks may require sensitive information of participating users to calculate solutions and decision variables, such as in traffic or energy systems. Adversaries with access to coordination signals may potentially decode information on individual users and put user privacy at risk. We develop local differential privacy, which is a stron… ▽ More Real-time data-driven optimization and control problems over networks may require sensitive information of participating users to calculate solutions and decision variables, such as in traffic or energy systems. Adversaries with access to coordination signals may potentially decode information on individual users and put user privacy at risk. We develop local differential privacy, which is a strong notion that guarantees user privacy regardless of any auxiliary information an adversary may have, for a larger family of convex distributed optimization problems. The mechanism allows agent to customize their own privacy level based on local needs and parameter sensitivities. We propose a general sampling based approach for determining sensitivity and derive analytical bounds for specific quadratic problems. We analyze inherent trade-offs between privacy and suboptimality and propose allocation schemes to divide the maximum allowable noise, a privacy budget, among all participating agents. Our algorithm is implemented to enable privacy in distributed optimal power flow for electric grids. △ Less

Submitted 22 May, 2020; v1 submitted 15 June, 2018; originally announced June 2018.

Comments: A shorter version of this paper will appear in the proceedings of ISGT-Europe 2020 in The Hague

Defending Against Saddle Point Attack in Byzantine-Robust Distributed Learning

Authors: Dong Yin, Yudong Chen, Kannan Ramchandran, Peter Bartlett

Abstract: We study robust distributed learning that involves minimizing a non-convex loss function with saddle points. We consider the Byzantine setting where some worker machines have abnormal or even arbitrary and adversarial behavior. In this setting, the Byzantine machines may create fake local minima near a saddle point that is far away from any true local minimum, even when robust gradient estimators… ▽ More We study robust distributed learning that involves minimizing a non-convex loss function with saddle points. We consider the Byzantine setting where some worker machines have abnormal or even arbitrary and adversarial behavior. In this setting, the Byzantine machines may create fake local minima near a saddle point that is far away from any true local minimum, even when robust gradient estimators are used. We develop ByzantinePGD, a robust first-order algorithm that can provably escape saddle points and fake local minima, and converge to an approximate true local minimizer with low iteration complexity. As a by-product, we give a simpler algorithm and analysis for esca** saddle points in the usual non-Byzantine setting. We further discuss three robust gradient estimators that can be used in ByzantinePGD, including median, trimmed mean, and iterative filtering. We characterize their performance in concrete statistical settings, and argue for their near-optimality in low and high dimensional regimes. △ Less

Submitted 29 July, 2020; v1 submitted 14 June, 2018; originally announced June 2018.

Comments: ICML 2019