-
Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations
Authors:
Rylan Schaeffer,
Victor Lecomte,
Dhruv Bhandarkar Pai,
Andres Carranza,
Berivan Isik,
Alyssa Unell,
Mikail Khona,
Thomas Yerxa,
Yann LeCun,
SueYeon Chung,
Andrey Gromov,
Ravid Shwartz-Ziv,
Sanmi Koyejo
Abstract:
Maximum Manifold Capacity Representations (MMCR) is a recent multi-view self-supervised learning (MVSSL) method that matches or surpasses other leading MVSSL methods. MMCR is intriguing because it does not fit neatly into any of the commonplace MVSSL lineages, instead originating from a statistical mechanical perspective on the linear separability of data manifolds. In this paper, we seek to impro…
▽ More
Maximum Manifold Capacity Representations (MMCR) is a recent multi-view self-supervised learning (MVSSL) method that matches or surpasses other leading MVSSL methods. MMCR is intriguing because it does not fit neatly into any of the commonplace MVSSL lineages, instead originating from a statistical mechanical perspective on the linear separability of data manifolds. In this paper, we seek to improve our understanding and our utilization of MMCR. To better understand MMCR, we leverage tools from high dimensional probability to demonstrate that MMCR incentivizes alignment and uniformity of learned embeddings. We then leverage tools from information theory to show that such embeddings maximize a well-known lower bound on mutual information between views, thereby connecting the geometric perspective of MMCR to the information-theoretic perspective commonly discussed in MVSSL. To better utilize MMCR, we mathematically predict and experimentally confirm non-monotonic changes in the pretraining loss akin to double descent but with respect to atypical hyperparameters. We also discover compute scaling laws that enable predicting the pretraining loss as a function of gradients steps, batch size, embedding dimension and number of views. We then show that MMCR, originally applied to image data, is performant on multimodal image-text data. By more deeply understanding the theoretical and empirical behavior of MMCR, our work reveals insights on improving MVSSL methods.
△ Less
Submitted 13 June, 2024;
originally announced June 2024.
-
Grokking Modular Polynomials
Authors:
Darshil Doshi,
Tianyu He,
Aritra Das,
Andrey Gromov
Abstract:
Neural networks readily learn a subset of the modular arithmetic tasks, while failing to generalize on the rest. This limitation remains unmoved by the choice of architecture and training strategies. On the other hand, an analytical solution for the weights of Multi-layer Perceptron (MLP) networks that generalize on the modular addition task is known in the literature. In this work, we (i) extend…
▽ More
Neural networks readily learn a subset of the modular arithmetic tasks, while failing to generalize on the rest. This limitation remains unmoved by the choice of architecture and training strategies. On the other hand, an analytical solution for the weights of Multi-layer Perceptron (MLP) networks that generalize on the modular addition task is known in the literature. In this work, we (i) extend the class of analytical solutions to include modular multiplication as well as modular addition with many terms. Additionally, we show that real networks trained on these datasets learn similar solutions upon generalization (grokking). (ii) We combine these "expert" solutions to construct networks that generalize on arbitrary modular polynomials. (iii) We hypothesize a classification of modular polynomials into learnable and non-learnable via neural networks training; and provide experimental evidence supporting our claims.
△ Less
Submitted 5 June, 2024;
originally announced June 2024.
-
Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks
Authors:
Tianyu He,
Darshil Doshi,
Aritra Das,
Andrey Gromov
Abstract:
Large language models can solve tasks that were not present in the training set. This capability is believed to be due to in-context learning and skill composition. In this work, we study the emergence of in-context learning and skill composition in a collection of modular arithmetic tasks. Specifically, we consider a finite collection of linear modular functions…
▽ More
Large language models can solve tasks that were not present in the training set. This capability is believed to be due to in-context learning and skill composition. In this work, we study the emergence of in-context learning and skill composition in a collection of modular arithmetic tasks. Specifically, we consider a finite collection of linear modular functions $z = a \, x + b \, y \;\mathrm{mod}\; p$ labeled by the vector $(a, b) \in \mathbb{Z}_p^2$. We use some of these tasks for pre-training and the rest for out-of-distribution testing. We empirically show that a GPT-style transformer exhibits a transition from in-distribution to out-of-distribution generalization as the number of pre-training tasks increases. We find that the smallest model capable of out-of-distribution generalization requires two transformer blocks, while for deeper models, the out-of-distribution generalization phase is \emph{transient}, necessitating early stop**. Finally, we perform an interpretability study of the pre-trained models, revealing the highly structured representations in both phases; and discuss the learnt algorithm.
△ Less
Submitted 4 June, 2024;
originally announced June 2024.
-
Is Model Collapse Inevitable? Breaking the Curse of Recursion by Accumulating Real and Synthetic Data
Authors:
Matthias Gerstgrasser,
Rylan Schaeffer,
Apratim Dey,
Rafael Rafailov,
Henry Sleight,
John Hughes,
Tomasz Korbak,
Rajashree Agrawal,
Dhruv Pai,
Andrey Gromov,
Daniel A. Roberts,
Diyi Yang,
David L. Donoho,
Sanmi Koyejo
Abstract:
The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration…
▽ More
The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration until fitted models become useless. However, those studies largely assumed that new data replace old data over time, where an arguably more realistic assumption is that data accumulate over time. In this paper, we ask: what effect does accumulating data have on model collapse? We empirically study this question by pretraining sequences of language models on text corpora. We confirm that replacing the original real data by each generation's synthetic data does indeed tend towards model collapse, then demonstrate that accumulating the successive generations of synthetic data alongside the original real data avoids model collapse; these results hold across a range of model sizes, architectures, and hyperparameters. We obtain similar results for deep generative models on other types of real data: diffusion models for molecule conformation generation and variational autoencoders for image generation. To understand why accumulating data can avoid model collapse, we use an analytically tractable framework introduced by prior work in which a sequence of linear models are fit to the previous models' outputs. Previous work used this framework to show that if data are replaced, the test error increases with the number of model-fitting iterations; we extend this argument to prove that if data instead accumulate, the test error has a finite upper bound independent of the number of iterations, meaning model collapse no longer occurs.
△ Less
Submitted 29 April, 2024; v1 submitted 1 April, 2024;
originally announced April 2024.
-
The Unreasonable Ineffectiveness of the Deeper Layers
Authors:
Andrey Gromov,
Kushal Tirumala,
Hassan Shapourian,
Paolo Glorioso,
Daniel A. Roberts
Abstract:
We empirically study a simple layer-pruning strategy for popular families of open-weight pretrained LLMs, finding minimal degradation of performance on different question-answering benchmarks until after a large fraction (up to half) of the layers are removed. To prune these models, we identify the optimal block of layers to prune by considering similarity across layers; then, to "heal" the damage…
▽ More
We empirically study a simple layer-pruning strategy for popular families of open-weight pretrained LLMs, finding minimal degradation of performance on different question-answering benchmarks until after a large fraction (up to half) of the layers are removed. To prune these models, we identify the optimal block of layers to prune by considering similarity across layers; then, to "heal" the damage, we perform a small amount of finetuning. In particular, we use parameter-efficient finetuning (PEFT) methods, specifically quantization and Low Rank Adapters (QLoRA), such that each of our experiments can be performed on a single A100 GPU. From a practical perspective, these results suggest that layer pruning methods can complement other PEFT strategies to further reduce computational resources of finetuning on the one hand, and can improve the memory and latency of inference on the other hand. From a scientific perspective, the robustness of these LLMs to the deletion of layers implies either that current pretraining methods are not properly leveraging the parameters in the deeper layers of the network or that the shallow layers play a critical role in storing knowledge.
△ Less
Submitted 26 March, 2024;
originally announced March 2024.
-
Spot the bot: Coarse-Grained Partition of Semantic Paths for Bots and Humans
Authors:
Vasilii A. Gromov,
Alexandra S. Kogan
Abstract:
Nowadays, technology is rapidly advancing: bots are writing comments, articles, and reviews. Due to this fact, it is crucial to know if the text was written by a human or by a bot. This paper focuses on comparing structures of the coarse-grained partitions of semantic paths for human-written and bot-generated texts. We compare the clusterizations of datasets of n-grams from literary texts and text…
▽ More
Nowadays, technology is rapidly advancing: bots are writing comments, articles, and reviews. Due to this fact, it is crucial to know if the text was written by a human or by a bot. This paper focuses on comparing structures of the coarse-grained partitions of semantic paths for human-written and bot-generated texts. We compare the clusterizations of datasets of n-grams from literary texts and texts generated by several bots. The hypothesis is that the structures and clusterizations are different. Our research supports the hypothesis. As the semantic structure may be different for different languages, we investigate Russian, English, German, and Vietnamese languages.
△ Less
Submitted 27 February, 2024;
originally announced February 2024.
-
Bridging Associative Memory and Probabilistic Modeling
Authors:
Rylan Schaeffer,
Nika Zahedi,
Mikail Khona,
Dhruv Pai,
Sang Truong,
Yilun Du,
Mitchell Ostrow,
Sarthak Chandra,
Andres Carranza,
Ila Rani Fiete,
Andrey Gromov,
Sanmi Koyejo
Abstract:
Associative memory and probabilistic modeling are two fundamental topics in artificial intelligence. The first studies recurrent neural networks designed to denoise, complete and retrieve data, whereas the second studies learning and sampling from probability distributions. Based on the observation that associative memory's energy functions can be seen as probabilistic modeling's negative log like…
▽ More
Associative memory and probabilistic modeling are two fundamental topics in artificial intelligence. The first studies recurrent neural networks designed to denoise, complete and retrieve data, whereas the second studies learning and sampling from probability distributions. Based on the observation that associative memory's energy functions can be seen as probabilistic modeling's negative log likelihoods, we build a bridge between the two that enables useful flow of ideas in both directions. We showcase four examples: First, we propose new energy-based models that flexibly adapt their energy functions to new in-context datasets, an approach we term \textit{in-context learning of energy functions}. Second, we propose two new associative memory models: one that dynamically creates new memories as necessitated by the training data using Bayesian nonparametrics, and another that explicitly computes proportional memory assignments using the evidence lower bound. Third, using tools from associative memory, we analytically and numerically characterize the memory capacity of Gaussian kernel density estimators, a widespread tool in probababilistic modeling. Fourth, we study a widespread implementation choice in transformers -- normalization followed by self attention -- to show it performs clustering on the hypersphere. Altogether, this work urges further exchange of useful ideas between these two continents of artificial intelligence.
△ Less
Submitted 13 June, 2024; v1 submitted 15 February, 2024;
originally announced February 2024.
-
Formal concept analysis for evaluating intrinsic dimension of a natural language
Authors:
Sergei O. Kuznetsov,
Vasilii A. Gromov,
Nikita S. Borodin,
Andrei M. Divavin
Abstract:
Some results of a computational experiment for determining the intrinsic dimension of linguistic varieties for the Bengali and Russian languages are presented. At the same time, both sets of words and sets of bigrams in these languages were considered separately. The method used to solve this problem was based on formal concept analysis algorithms. It was found that the intrinsic dimensions of the…
▽ More
Some results of a computational experiment for determining the intrinsic dimension of linguistic varieties for the Bengali and Russian languages are presented. At the same time, both sets of words and sets of bigrams in these languages were considered separately. The method used to solve this problem was based on formal concept analysis algorithms. It was found that the intrinsic dimensions of these languages are significantly less than the dimensions used in popular neural network models in natural language processing.
△ Less
Submitted 17 November, 2023;
originally announced November 2023.
-
A Language and Its Dimensions: Intrinsic Dimensions of Language Fractal Structures
Authors:
Vasilii A. Gromov,
Nikita S. Borodin,
Asel S. Yerbolova
Abstract:
The present paper introduces a novel object of study - a language fractal structure. We hypothesize that a set of embeddings of all $n$-grams of a natural language constitutes a representative sample of this fractal set. (We use the term Hailonakea to refer to the sum total of all language fractal structures, over all $n$). The paper estimates intrinsic (genuine) dimensions of language fractal str…
▽ More
The present paper introduces a novel object of study - a language fractal structure. We hypothesize that a set of embeddings of all $n$-grams of a natural language constitutes a representative sample of this fractal set. (We use the term Hailonakea to refer to the sum total of all language fractal structures, over all $n$). The paper estimates intrinsic (genuine) dimensions of language fractal structures for the Russian and English languages. To this end, we employ methods based on (1) topological data analysis and (2) a minimum spanning tree of a data graph for a cloud of points considered (Steele theorem). For both languages, for all $n$, the intrinsic dimensions appear to be non-integer values (typical for fractal sets), close to 9 for both of the Russian and English language.
△ Less
Submitted 20 November, 2023; v1 submitted 16 November, 2023;
originally announced November 2023.
-
To grok or not to grok: Disentangling generalization and memorization on corrupted algorithmic datasets
Authors:
Darshil Doshi,
Aritra Das,
Tianyu He,
Andrey Gromov
Abstract:
Robust generalization is a major challenge in deep learning, particularly when the number of trainable parameters is very large. In general, it is very difficult to know if the network has memorized a particular set of examples or understood the underlying rule (or both). Motivated by this challenge, we study an interpretable model where generalizing representations are understood analytically, an…
▽ More
Robust generalization is a major challenge in deep learning, particularly when the number of trainable parameters is very large. In general, it is very difficult to know if the network has memorized a particular set of examples or understood the underlying rule (or both). Motivated by this challenge, we study an interpretable model where generalizing representations are understood analytically, and are easily distinguishable from the memorizing ones. Namely, we consider multi-layer perceptron (MLP) and Transformer architectures trained on modular arithmetic tasks, where ($ξ\cdot 100\%$) of labels are corrupted (\emph{i.e.} some results of the modular operations in the training set are incorrect). We show that (i) it is possible for the network to memorize the corrupted labels \emph{and} achieve $100\%$ generalization at the same time; (ii) the memorizing neurons can be identified and pruned, lowering the accuracy on corrupted data and improving the accuracy on uncorrupted data; (iii) regularization methods such as weight decay, dropout and BatchNorm force the network to ignore the corrupted data during optimization, and achieve $100\%$ accuracy on the uncorrupted dataset; and (iv) the effect of these regularization methods is (``mechanistically'') interpretable: weight decay and dropout force all the neurons to learn generalizing representations, while BatchNorm de-amplifies the output of memorizing neurons and amplifies the output of the generalizing ones. Finally, we show that in the presence of regularization, the training dynamics involves two consecutive stages: first, the network undergoes \emph{grokking} dynamics reaching high train \emph{and} test accuracy; second, it unlearns the memorizing representations, where the train accuracy suddenly jumps from $100\%$ to $100 (1-ξ)\%$.
△ Less
Submitted 4 March, 2024; v1 submitted 19 October, 2023;
originally announced October 2023.
-
Genetic Engineering Algorithm (GEA): An Efficient Metaheuristic Algorithm for Solving Combinatorial Optimization Problems
Authors:
Majid Sohrabi,
Amir M. Fathollahi-Fard,
Vasilii A. Gromov
Abstract:
Genetic Algorithms (GAs) are known for their efficiency in solving combinatorial optimization problems, thanks to their ability to explore diverse solution spaces, handle various representations, exploit parallelism, preserve good solutions, adapt to changing dynamics, handle combinatorial diversity, and provide heuristic search. However, limitations such as premature convergence, lack of problem-…
▽ More
Genetic Algorithms (GAs) are known for their efficiency in solving combinatorial optimization problems, thanks to their ability to explore diverse solution spaces, handle various representations, exploit parallelism, preserve good solutions, adapt to changing dynamics, handle combinatorial diversity, and provide heuristic search. However, limitations such as premature convergence, lack of problem-specific knowledge, and randomness of crossover and mutation operators make GAs generally inefficient in finding an optimal solution. To address these limitations, this paper proposes a new metaheuristic algorithm called the Genetic Engineering Algorithm (GEA) that draws inspiration from genetic engineering concepts. GEA redesigns the traditional GA while incorporating new search methods to isolate, purify, insert, and express new genes based on existing ones, leading to the emergence of desired traits and the production of specific chromosomes based on the selected genes. Comparative evaluations against state-of-the-art algorithms on benchmark instances demonstrate the superior performance of GEA, showcasing its potential as an innovative and efficient solution for combinatorial optimization problems.
△ Less
Submitted 28 September, 2023;
originally announced September 2023.
-
Date-Driven Approach for Identifying State of Hemodialysis Fistulas: Entropy-Complexity and Formal Concept Analysis
Authors:
Vasilii A. Gromov,
E. I. Zvorykina,
Yurii N. Beschastnov,
Majid Sohrabi
Abstract:
The paper explores mathematical methods that differentiate regular and chaotic time series, specifically for identifying pathological fistulas. It proposes a noise-resistant method for classifying responding rows of normally and pathologically functioning fistulas. This approach is grounded in the hypothesis that laminar blood flow signifies normal function, while turbulent flow indicates patholog…
▽ More
The paper explores mathematical methods that differentiate regular and chaotic time series, specifically for identifying pathological fistulas. It proposes a noise-resistant method for classifying responding rows of normally and pathologically functioning fistulas. This approach is grounded in the hypothesis that laminar blood flow signifies normal function, while turbulent flow indicates pathology. The study explores two distinct methods for distinguishing chaotic from regular time series. The first method involves map** the time series onto the entropy-complexity plane and subsequently comparing it to established clusters. The second method, introduced by the authors, constructs a concepts-objects graph using formal concept analysis. Both of these methods exhibit high efficiency in determining the state of the fistula.
△ Less
Submitted 25 September, 2023;
originally announced September 2023.
-
Signatures of Supersymmetry in the $ν{=}5/2$ Fractional Quantum Hall Effect
Authors:
Songyang Pu,
Ajit C. Balram,
Mikael Fremling,
Andrey Gromov,
Zlatko Papić
Abstract:
The Moore-Read state, one of the leading candidates for describing the fractional quantum Hall effect at filling factor $ν{=}5/2$, is a paradigmatic $p$-wave superconductor with non-Abelian topological order. Among its many exotic properties, the state hosts two collective modes: a bosonic density wave and a neutral fermion mode that arises from an unpaired electron in the condensate. It has recen…
▽ More
The Moore-Read state, one of the leading candidates for describing the fractional quantum Hall effect at filling factor $ν{=}5/2$, is a paradigmatic $p$-wave superconductor with non-Abelian topological order. Among its many exotic properties, the state hosts two collective modes: a bosonic density wave and a neutral fermion mode that arises from an unpaired electron in the condensate. It has recently been proposed that the descriptions of the two modes can be unified by postulating supersymmetry (SUSY) that relates them in the long-wavelength limit. Here we extend the SUSY description to construct wave functions of the two modes on closed surfaces, such as the sphere and torus, and we test the resulting states in large-scale numerical simulations. We demonstrate the equivalence in the long-wavelength limit between SUSY wave functions and previous descriptions of collective modes based on the Girvin-MacDonald-Platzman ansatz, Jack polynomials, and bipartite composite fermions. Leveraging the first-quantized form of the SUSY wave functions, we study their energies using the Monte Carlo method and show that realistic $ν{=}5/2$ systems are close to the putative SUSY point, where the two collective modes become degenerate in energy.
△ Less
Submitted 7 May, 2023; v1 submitted 10 January, 2023;
originally announced January 2023.
-
Grokking modular arithmetic
Authors:
Andrey Gromov
Abstract:
We present a simple neural network that can learn modular arithmetic tasks and exhibits a sudden jump in generalization known as ``grokking''. Concretely, we present (i) fully-connected two-layer networks that exhibit grokking on various modular arithmetic tasks under vanilla gradient descent with the MSE loss function in the absence of any regularization; (ii) evidence that grokking modular arith…
▽ More
We present a simple neural network that can learn modular arithmetic tasks and exhibits a sudden jump in generalization known as ``grokking''. Concretely, we present (i) fully-connected two-layer networks that exhibit grokking on various modular arithmetic tasks under vanilla gradient descent with the MSE loss function in the absence of any regularization; (ii) evidence that grokking modular arithmetic corresponds to learning specific feature maps whose structure is determined by the task; (iii) analytic expressions for the weights -- and thus for the feature maps -- that solve a large class of modular arithmetic tasks; and (iv) evidence that these feature maps are also found by vanilla gradient descent as well as AdamW, thereby establishing complete interpretability of the representations learnt by the network.
△ Less
Submitted 6 January, 2023;
originally announced January 2023.
-
Supergravity model of the Haldane-Rezayi fractional quantum Hall state
Authors:
Dung Xuan Nguyen,
Kartik Prabhu,
Ajit C. Balram,
Andrey Gromov
Abstract:
Supersymmetry and supergravity were invented in the 1970s to solve fundamental problems in high-energy physics. Even though neither of these ideas has yet been confirmed in high-energy and cosmology experiments, they have been beneficial in constructing numerous theoretical models, including superstring theory. Despite the absence of supersymmetry in particle physics, it can potentially emerge in…
▽ More
Supersymmetry and supergravity were invented in the 1970s to solve fundamental problems in high-energy physics. Even though neither of these ideas has yet been confirmed in high-energy and cosmology experiments, they have been beneficial in constructing numerous theoretical models, including superstring theory. Despite the absence of supersymmetry in particle physics, it can potentially emerge in exotic phases of strongly correlated condensed matter systems. In this paper, we propose a supergravity model that describes the low-energy physics of the Haldane-Rezayi state, a gapless quantum Hall state that occurs in a half-filled Landau level. We show that the corresponding edge modes of the Haldane-Rezayi state and the Girvin-MacDonald-Platzman algebra appear naturally in the supergravity model. Finally, we substantiate our theoretical findings with numerical exact diagonalization calculations that support the appearance of the emergent graviton and gravitino excitations in the Haldane-Rezayi state.
△ Less
Submitted 9 March, 2023; v1 submitted 1 December, 2022;
originally announced December 2022.
-
Fracton Matter
Authors:
Andrey Gromov,
Leo Radzihovsky
Abstract:
We review a burgeoning field of "fractons" -- a class of models where quasi-particles are strictly immobile or display restricted mobility that can be understood through generalized multipolar symmetries and associated conservation laws. Focusing on just a corner of this fast-growing subject, we will demonstrate how one class of such theories -- symmetric tensor and coupled-vector gauge theories s…
▽ More
We review a burgeoning field of "fractons" -- a class of models where quasi-particles are strictly immobile or display restricted mobility that can be understood through generalized multipolar symmetries and associated conservation laws. Focusing on just a corner of this fast-growing subject, we will demonstrate how one class of such theories -- symmetric tensor and coupled-vector gauge theories surprisingly emerge from familiar elasticity of a two-dimensional quantum crystal. The disclination and dislocation crystal defects respectively map onto charges and dipoles of the fracton gauge theory. This fracton-elasticity duality leads to predictions of fractonic phases and quantum phase transitions to their descendants, that are duals of the commensurate crystal, supersolid, smectic, hexatic liquid crystals, as well as amorphous solids, quasi-crystals and elastic membranes. We show how these dual gauge theories provide a field theoretic description of quantum melting transitions through a generalized Higgs mechanism. We demonstrate how they can be equivalently constructed as gauged models with global multipole symmetries. We expect extensions of such gauge-elasticity dualities to generalized elasticity theories provide a route to discovery of new fractonic models and their potential experimental realizations.
△ Less
Submitted 9 November, 2022;
originally announced November 2022.
-
The contraction hypothesis of the gauge group of the Standard Model and LHC experimental data
Authors:
N. A. Gromov
Abstract:
Within the framework of the contraction hypothesis of the gauge group of the Standard Model the behavior of the amplitude of the dominant Higgs boson production process in the four-lepton decay with increasing temperature $T$ is analyzed. It is shown that the modified process breaks down into a number of channels depending on the contribution of the color components in the loop of virtual quarks,…
▽ More
Within the framework of the contraction hypothesis of the gauge group of the Standard Model the behavior of the amplitude of the dominant Higgs boson production process in the four-lepton decay with increasing temperature $T$ is analyzed. It is shown that the modified process breaks down into a number of channels depending on the contribution of the color components in the loop of virtual quarks, leading to the creation of the Higgs boson. The dependence on $T$ of the cross section of each channel is found. Comparison with LHC data on Higgs boson creation cross sections at energies (temperatures) of 7, 8, 13, and 14 TeV showed that the hypothesis about the contraction of the gauge group of the Standard Model does not contradict these data.
△ Less
Submitted 7 September, 2022;
originally announced September 2022.
-
AutoInit: Automatic Initialization via Jacobian Tuning
Authors:
Tianyu He,
Darshil Doshi,
Andrey Gromov
Abstract:
Good initialization is essential for training Deep Neural Networks (DNNs). Oftentimes such initialization is found through a trial and error approach, which has to be applied anew every time an architecture is substantially modified, or inherited from smaller size networks leading to sub-optimal initialization. In this work we introduce a new and cheap algorithm, that allows one to find a good ini…
▽ More
Good initialization is essential for training Deep Neural Networks (DNNs). Oftentimes such initialization is found through a trial and error approach, which has to be applied anew every time an architecture is substantially modified, or inherited from smaller size networks leading to sub-optimal initialization. In this work we introduce a new and cheap algorithm, that allows one to find a good initialization automatically, for general feed-forward DNNs. The algorithm utilizes the Jacobian between adjacent network blocks to tune the network hyperparameters to criticality. We solve the dynamics of the algorithm for fully connected networks with ReLU and derive conditions for its convergence. We then extend the discussion to more general architectures with BatchNorm and residual connections. Finally, we apply our method to ResMLP and VGG architectures, where the automatic one-shot initialization found by our method shows good performance on vision tasks.
△ Less
Submitted 27 June, 2022;
originally announced June 2022.
-
Critical Initialization of Wide and Deep Neural Networks through Partial Jacobians: General Theory and Applications
Authors:
Darshil Doshi,
Tianyu He,
Andrey Gromov
Abstract:
Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to infinity, the network function is a Gaussian process (GP) and quantitatively predictive description is possible. Gaussian approximation allows one to formulate criteria for selecting hyperparameters, such as variances of weights and biases, as well as the learning rat…
▽ More
Deep neural networks are notorious for defying theoretical treatment. However, when the number of parameters in each layer tends to infinity, the network function is a Gaussian process (GP) and quantitatively predictive description is possible. Gaussian approximation allows one to formulate criteria for selecting hyperparameters, such as variances of weights and biases, as well as the learning rate. These criteria rely on the notion of criticality defined for deep neural networks. In this work we describe a new practical way to diagnose criticality. We introduce \emph{partial Jacobians} of a network, defined as derivatives of preactivations in layer $l$ with respect to preactivations in layer $l_0\leq l$. We derive recurrence relations for the norms of partial Jacobians and utilize these relations to analyze criticality of deep fully connected neural networks with LayerNorm and/or residual connections. We derive and implement a simple and cheap numerical test that allows one to select optimal initialization for a broad class of deep neural networks; containing fully connected, convolutional and normalization layers. Using these tools we show quantitatively that proper stacking of the LayerNorm (applied to preactivations) and residual connections leads to an architecture that is critical for any initialization. Finally, we apply our methods to analyze ResNet and MLP-Mixer architectures; demonstrating the everywhere-critical regime.
△ Less
Submitted 5 October, 2023; v1 submitted 23 November, 2021;
originally announced November 2021.
-
Very high-energy collective states of partons in fractional quantum Hall liquids
Authors:
Ajit C. Balram,
Zhao Liu,
Andrey Gromov,
Zlatko Papić
Abstract:
The low energy physics of fractional quantum Hall (FQH) states -- a paradigm of strongly correlated topological phases of matter -- to a large extent is captured by weakly interacting quasiparticles known as composite fermions (CFs). In this paper, based on numerical simulations and effective field theory, we argue that some \emph{high energy} states in the FQH spectra necessitate a different desc…
▽ More
The low energy physics of fractional quantum Hall (FQH) states -- a paradigm of strongly correlated topological phases of matter -- to a large extent is captured by weakly interacting quasiparticles known as composite fermions (CFs). In this paper, based on numerical simulations and effective field theory, we argue that some \emph{high energy} states in the FQH spectra necessitate a different description based on \emph{parton} quasiparticles. We show that Jain states at filling factor $ν{=}n/(2pn\pm1)$ with integers $n,p{\geq}2$, support two kinds of collective modes: in addition to the well-known Girvin-MacDonald-Platzman (GMP) mode, they host a high energy collective mode, which is interpreted as the GMP mode of partons. We elucidate observable signatures of the parton mode in the dynamics following a geometric quench. We construct a microscopic wave function for the parton mode, and demonstrate agreement between its variational energy and exact diagonalization. Using the parton construction, we derive a field theory of the Jain states and show that the previously proposed effective theories follow from our approach. Our results point to partons being "real" quasiparticles which, in a way reminiscent of quarks, only become observable at sufficiently high energies.
△ Less
Submitted 12 April, 2022; v1 submitted 19 November, 2021;
originally announced November 2021.
-
Three-component Stackel model of the Galaxy based on the rotation curve from maser data
Authors:
A. O. Gromov,
I. I. Nikiforov
Abstract:
A three-component Stackel model of the Galaxy, including the bulge, disk, and halo, is constructed. Parameter estimates of the potential are obtained as a result of fitting the model rotation curve to azimuthal velocities found from data on trigonometric parallaxes and spatial velocities of masers. The fitting method takes into account the measurement and natural dispersions of azimuthal velocitie…
▽ More
A three-component Stackel model of the Galaxy, including the bulge, disk, and halo, is constructed. Parameter estimates of the potential are obtained as a result of fitting the model rotation curve to azimuthal velocities found from data on trigonometric parallaxes and spatial velocities of masers. The fitting method takes into account the measurement and natural dispersions of azimuthal velocities and uses an algorithm for excluding objects with excessive residuals. In order to obtain more uniform samples, the objects were divided into two groups: masers associated with high-mass star forming regions and masers of other types. A significant kinematic inhomogeneity of these groups was identified and taken into account: the azimuthal velocity dispersion is $σ_{0,1}=4.3\pm 0.4$~km\,s$^{-1}$, in the first group and $σ_{0,2}=15.2\pm1.3$~km\,s$^{-1}$ in the second. After constructing the model of the Galactic-plane potential, it was generalized to the entire space under the assumption of the existence of a third quadratic integral of motion. When reconstructing the Galactic rotation curve in detail, the used algorithm gives an analytical expression for the Stackel potential, which significantly simplifies the task of constructing the Galaxy's phase density model in the Stackel approximation. In order to make the Stackel model more realistic, one needs to develop methods of direct account of data on the vertical distribution of density in the Galaxy.
△ Less
Submitted 23 October, 2021;
originally announced October 2021.
-
Particle properties in the early universe from the contraction of the SM gauge group
Authors:
Nikolai A. Gromov
Abstract:
The properties of elementary particles and their interactions at different stages of the evolution of the Universe, starting with the Planck energy $ 10 ^{19} $ GeV, are presented. We assume that the Standard Model gauge group becomes simpler as the temperature of the universe increases. The description is based on the hypothesis that the high-energy (high-temperature) limit of the Standard Model…
▽ More
The properties of elementary particles and their interactions at different stages of the evolution of the Universe, starting with the Planck energy $ 10 ^{19} $ GeV, are presented. We assume that the Standard Model gauge group becomes simpler as the temperature of the universe increases. The description is based on the hypothesis that the high-energy (high-temperature) limit of the Standard Model is generated by the contraction of the gauge group. An explicit form of the Lagrangian is obtained for each stage of the evolution of the universe and is the basis for describing the properties of elementary particles. These properties change drastically in the infinite temperature limit: all particles lose mass, only massless neutral $ Z $ bosons and $ u $ quarks, as well as neutrinos and photons, survive. Electroweak interactions become long range and are mediated by neutral currents. All quarks are monochromatic.
△ Less
Submitted 25 August, 2021;
originally announced August 2021.
-
Higher-Energy Standard Model from the Gauge Group Contraction
Authors:
N. A. Gromov
Abstract:
The evolution of properties and interactions of elementary particles is described, beginning with the Planck scale of $10^{19}$ GeV. The description is based on the hypothesis that high-temperature (high-energy) limit of the Standard Model is generated by the gauge group contraction. In the infinite-temperature limit, properties of particles fundamentally change: all particles lose their masses, o…
▽ More
The evolution of properties and interactions of elementary particles is described, beginning with the Planck scale of $10^{19}$ GeV. The description is based on the hypothesis that high-temperature (high-energy) limit of the Standard Model is generated by the gauge group contraction. In the infinite-temperature limit, properties of particles fundamentally change: all particles lose their masses, only massless neutral $Z$ bosons and $u$ quarks together with neutrinos and photons survive. Weak interactions become long-range ones and are generated by neutral currents. Quarks have only one color degree of freedom.
△ Less
Submitted 28 May, 2021;
originally announced May 2021.
-
Quantum many-body topology of quasicrystals
Authors:
Dominic V. Else,
Sheng-Jie Huang,
Abhinav Prem,
Andrey Gromov
Abstract:
In this paper, we characterize quasicrystalline interacting topological phases of matter i.e., phases protected by some quasicrystalline structure. We show that the elasticity theory of quasicrystals, which accounts for both "phonon" and "phason" modes, admits non-trivial quantized topological terms with far richer structure than their crystalline counterparts. We show that these terms correspond…
▽ More
In this paper, we characterize quasicrystalline interacting topological phases of matter i.e., phases protected by some quasicrystalline structure. We show that the elasticity theory of quasicrystals, which accounts for both "phonon" and "phason" modes, admits non-trivial quantized topological terms with far richer structure than their crystalline counterparts. We show that these terms correspond to distinct phases of matter and also uncover intrinsically quasicrystalline phases, which have no crystalline analogues. For quasicrystals with internal $\mathrm{U}(1)$ symmetry, we discuss a number of interpretations and physical implications of the topological terms, including constraints on the mobility of dislocations in $d=2$ quasicrystals and a quasicrystalline generalization of the Lieb-Schultz-Mattis-Oshikawa-Hastings theorem. We then extend these ideas much further and address the complete classification of quasicrystalline topological phases, including systems with point-group symmetry as well as non-invertible phases. We hence obtain the "Quasicrystalline Equivalence Principle," which generalizes the classification of crystalline topological phases to the quasicrystalline setting.
△ Less
Submitted 12 April, 2021; v1 submitted 24 March, 2021;
originally announced March 2021.
-
Conceptual design of the Spin Physics Detector
Authors:
V. M. Abazov,
V. Abramov,
L. G. Afanasyev,
R. R. Akhunzyanov,
A. V. Akindinov,
N. Akopov,
I. G. Alekseev,
A. M. Aleshko,
V. Yu. Alexakhin,
G. D. Alexeev,
M. Alexeev,
A. Amoroso,
I. V. Anikin,
V. F. Andreev,
V. A. Anosov,
A. B. Arbuzov,
N. I. Azorskiy,
A. A. Baldin,
V. V. Balandina,
E. G. Baldina,
M. Yu. Barabanov,
S. G. Barsov,
V. A. Baskov,
A. N. Beloborodov,
I. N. Belov
, et al. (270 additional authors not shown)
Abstract:
The Spin Physics Detector, a universal facility for studying the nucleon spin structure and other spin-related phenomena with polarized proton and deuteron beams, is proposed to be placed in one of the two interaction points of the NICA collider that is under construction at the Joint Institute for Nuclear Research (Dubna, Russia). At the heart of the project there is huge experience with polarize…
▽ More
The Spin Physics Detector, a universal facility for studying the nucleon spin structure and other spin-related phenomena with polarized proton and deuteron beams, is proposed to be placed in one of the two interaction points of the NICA collider that is under construction at the Joint Institute for Nuclear Research (Dubna, Russia). At the heart of the project there is huge experience with polarized beams at JINR.
The main objective of the proposed experiment is the comprehensive study of the unpolarized and polarized gluon content of the nucleon. Spin measurements at the Spin Physics Detector at the NICA collider have bright perspectives to make a unique contribution and challenge our understanding of the spin structure of the nucleon. In this document the Conceptual Design of the Spin Physics Detector is presented.
△ Less
Submitted 2 February, 2022; v1 submitted 31 January, 2021;
originally announced February 2021.
-
Quench dynamics of collective modes in fractional quantum Hall bilayers
Authors:
Zhao Liu,
Ajit C. Balram,
Zlatko Papić,
Andrey Gromov
Abstract:
We introduce different types of quenches to probe the non-equilibrium dynamics and multiple collective modes of bilayer fractional quantum Hall states. We show that applying an electric field in one layer induces oscillations of a spin-1 degree of freedom, whose frequency matches the long-wavelength limit of the dipole mode. On the other hand, oscillations of the long-wavelength limit of the quadr…
▽ More
We introduce different types of quenches to probe the non-equilibrium dynamics and multiple collective modes of bilayer fractional quantum Hall states. We show that applying an electric field in one layer induces oscillations of a spin-1 degree of freedom, whose frequency matches the long-wavelength limit of the dipole mode. On the other hand, oscillations of the long-wavelength limit of the quadrupole mode, i.e., the spin-2 graviton, as well as the combination of two spin-1 states, can be activated by a sudden change of band mass anisotropy. We construct an effective field theory to describe the quench dynamics of these collective modes. In particular, we derive the dynamics for both the spin-2 and the spin-1 states and demonstrate their excellent agreement with numerics.
△ Less
Submitted 3 November, 2020;
originally announced November 2020.
-
Fracton-elasticity duality of two-dimensional superfluid vortex crystals: defect interactions and quantum melting
Authors:
Dung Xuan Nguyen,
Andrey Gromov,
Sergej Moroz
Abstract:
Employing the fracton-elastic duality, we develop a low-energy effective theory of a zero-temperature vortex crystal in a two-dimensional bosonic superfluid which naturally incorporates crystalline topological defects. We extract static interactions between these defects and investigate several continuous quantum transitions triggered by the Higgs condensation of vortex vacancies/interstitials and…
▽ More
Employing the fracton-elastic duality, we develop a low-energy effective theory of a zero-temperature vortex crystal in a two-dimensional bosonic superfluid which naturally incorporates crystalline topological defects. We extract static interactions between these defects and investigate several continuous quantum transitions triggered by the Higgs condensation of vortex vacancies/interstitials and dislocations. We propose that the quantum melting of the vortex crystal towards the hexatic or smectic phase may occur via a pair of continuous transitions separated by an intermediate vortex supersolid phase.
△ Less
Submitted 9 November, 2020; v1 submitted 25 May, 2020;
originally announced May 2020.
-
Vortices and Fractons
Authors:
Darshil Doshi,
Andrey Gromov
Abstract:
We discuss a simple and experimentally available realization of fracton physics. We note that superfluid vortices form a Hamiltonian system that conserves total dipole moment and trace of the quadrupole moment of vorticity; thereby establishing a relation to a traceless scalar charge theory in two spatial dimensions. Next we consider the limit where the number of vortices is large and show that em…
▽ More
We discuss a simple and experimentally available realization of fracton physics. We note that superfluid vortices form a Hamiltonian system that conserves total dipole moment and trace of the quadrupole moment of vorticity; thereby establishing a relation to a traceless scalar charge theory in two spatial dimensions. Next we consider the limit where the number of vortices is large and show that emergent vortex hydrodynamics also conserves these moments. Finally, we show the motion of vortices and of fractons on curved surfaces agree, thereby opening a route to experimental study of the interplay between fracton physics and curved space. Our conclusions also apply to charged particles in strong magnetic field.
△ Less
Submitted 6 May, 2020;
originally announced May 2020.
-
Fracton hydrodynamics
Authors:
Andrey Gromov,
Andrew Lucas,
Rahul M. Nandkishore
Abstract:
We introduce new classes of hydrodynamic theories inspired by the recently discovered fracton phases of quantum matter. Fracton phases are characterized by elementary excitations (fractons) with restricted mobility. The hydrodynamic theories we introduce describe thermalization in systems with fracton-like mobility constraints, including fluids where charge and dipole moment are both locally conse…
▽ More
We introduce new classes of hydrodynamic theories inspired by the recently discovered fracton phases of quantum matter. Fracton phases are characterized by elementary excitations (fractons) with restricted mobility. The hydrodynamic theories we introduce describe thermalization in systems with fracton-like mobility constraints, including fluids where charge and dipole moment are both locally conserved, and fluids where charge is conserved along every line or plane of a lattice. Each of these fluids is subdiffusive, and constitutes a new universality class of hydrodynamic behavior. There are infinitely many such classes, each with distinct subdiffusive exponents, all of which are captured by our formalism. Our framework naturally explains recent results on dynamics with constrained quantum circuits, as well as recent experiments with ultracold atoms in tilted optical lattices. We identify crisp experimental signatures of these novel hydrodynamics, and explain how they may be realized in near term ultracold atom experiments.
△ Less
Submitted 22 July, 2020; v1 submitted 20 March, 2020;
originally announced March 2020.
-
Nonrelativistic quantum particles on the Minkowski plane
Authors:
N. A. Gromov,
I. V. Kostyakov,
V. V. Kuratov
Abstract:
The quantum-mechanical problems of a nonrelativistic free particle, a harmonic oscillator and a Coulomb particle on Minkowski plane are discussed. The Schrödinger equations for eigenvalues are obtained using the Beltrami-Laplas operator of the pseudo-Euclidean plane and the corresponding potentials. It is shown that, in contrast to the standard problem on Euclidean plane, in addition to the contin…
▽ More
The quantum-mechanical problems of a nonrelativistic free particle, a harmonic oscillator and a Coulomb particle on Minkowski plane are discussed. The Schrödinger equations for eigenvalues are obtained using the Beltrami-Laplas operator of the pseudo-Euclidean plane and the corresponding potentials. It is shown that, in contrast to the standard problem on Euclidean plane, in addition to the continuous spectrum, a free particle has a discrete energy levels and a Coulomb particle, in addition to the discrete spectrum, has unstable states that describe the incidence of a particle on isotropic lines forming a metric cone.
△ Less
Submitted 23 May, 2020; v1 submitted 27 February, 2020;
originally announced February 2020.
-
A Duality Between U(1) Haah Code and 3D Smectic A Phase
Authors:
Andrey Gromov
Abstract:
We describe a duality between multipole gauge theories and spatially ordered phases. Our main example is a duality between the multipole gauge theory description of the U(1) Haah code and smectic A phase in three spatial dimensions. We show how multipole symmetries restrict the mobility of dislocations and disclinations in smectic A phase. Finally, we exhibit a 2D version of the duality.
We describe a duality between multipole gauge theories and spatially ordered phases. Our main example is a duality between the multipole gauge theory description of the U(1) Haah code and smectic A phase in three spatial dimensions. We show how multipole symmetries restrict the mobility of dislocations and disclinations in smectic A phase. Finally, we exhibit a 2D version of the duality.
△ Less
Submitted 26 February, 2020;
originally announced February 2020.
-
Discovery of Two Types of Twinkling Can Explain Contradictory Observations among Twinkling Artifact Investigators in Ultrasound Imaging
Authors:
Denis Leonov,
Nicholas Kulberg,
Alexandr Gromov,
Anton Vladzymyrskyy,
Sergey Morozov
Abstract:
Twinkling Artifact is a valuable tool in detecting dense objects such as kidney stones, calculi etc., especially when there is no acoustic shadowing and presence of hyperechogenic tissues obstructs visualization. This phenomenon is not completely understood. Different scientific groups have contradictory findings concerning its properties: some of them observed a decrease in Twinkling intensity at…
▽ More
Twinkling Artifact is a valuable tool in detecting dense objects such as kidney stones, calculi etc., especially when there is no acoustic shadowing and presence of hyperechogenic tissues obstructs visualization. This phenomenon is not completely understood. Different scientific groups have contradictory findings concerning its properties: some of them observed a decrease in Twinkling intensity at elevated pulse repetition frequencies (PRF), while others found Twinkling to be independent from PRF, etc. In this paper we hypothesize that this kind of contradictions can be partially resolved on an assumption that there are two types of Twinkling. The 1st type presumably is produced by random-phased reflections from microcavitation and can be registered even at PRF high enough to suppress blood flow. The 2nd is originated from elastic vibrations as the object under investigation swings like a pendulum in the field of ultrasound. These vibrations can be associated with an external source such as a muscles contraction, etc. or the acoustic radiation force from signals emitted by the transducer. The 2nd type of Twinkling disappears at high PRF and can be regularly observed in silicone and polyurethane phantoms where the occurrence of cavitation microbubbles is highly unlikely. Key Words: Ultrasound Imaging, Twinkling Artifact, Stone Detection, Doppler Effect, Color Frame Map**, Acoustic Radiation Force, Elastic Vibration, Ultrasound Phantom, Cavitation, Microbubbles.
△ Less
Submitted 5 December, 2019;
originally announced December 2019.
-
Anisotropic odd viscosity via time-modulated drive
Authors:
Anton Souslov,
Andrey Gromov,
Vincenzo Vitelli
Abstract:
At equilibrium, the structure and response of ordered phases are typically determined by the spontaneous breaking of spatial symmetries. Out of equilibrium, spatial order itself can become a dynamically emergent concept. In this article, we show that spatially anisotropic viscous coefficients and stresses can be designed in a far-from-equilibrium fluid by applying to its constituents a time-modula…
▽ More
At equilibrium, the structure and response of ordered phases are typically determined by the spontaneous breaking of spatial symmetries. Out of equilibrium, spatial order itself can become a dynamically emergent concept. In this article, we show that spatially anisotropic viscous coefficients and stresses can be designed in a far-from-equilibrium fluid by applying to its constituents a time-modulated drive. If the drive induces a rotation whose rate is slowed down when the constituents point along specific directions, anisotropic structures and mechanical responses arise at long timescales. We demonstrate that the viscous response of such anisotropic driven fluids can acquire a tensorial, dissipationless component called anisotropic odd (or Hall) viscosity. Classical fluids with internal torques can display additional components of the odd viscosity neglected in previous studies of quantum Hall fluids that assumed angular momentum conservation. We show that these anisotropic and angular momentum-violating odd-viscosity coefficients can change even the bulk flow of an incompressible fluid by acting as a source of vorticity. In addition, shear distortions in the shape of an inclusion result in torques.
△ Less
Submitted 18 September, 2019;
originally announced September 2019.
-
Collective excitations at filling factor 5/2: The view from superspace
Authors:
Andrey Gromov,
Emil J. Martinec,
Shinsei Ryu
Abstract:
We present a microscopic theory of the neutral collective modes supported by the non-Abelian fractional quantum Hall states at filling factor 5/2. The theory is formulated in terms of the trial states describing the Girvin-MacDonald-Platzman (GMP) mode and its fermionic counterpart. These modes are superpartners of each other in a concrete sense, which we elucidate.
We present a microscopic theory of the neutral collective modes supported by the non-Abelian fractional quantum Hall states at filling factor 5/2. The theory is formulated in terms of the trial states describing the Girvin-MacDonald-Platzman (GMP) mode and its fermionic counterpart. These modes are superpartners of each other in a concrete sense, which we elucidate.
△ Less
Submitted 13 September, 2019;
originally announced September 2019.
-
On duality between Cosserat elasticity and fractons
Authors:
Andrey Gromov,
Piotr Surówka
Abstract:
We present a dual formulation of the Cosserat theory of elasticity. In this theory a local element of an elastic body is described in terms of local displacement and local orientation. Upon the duality transformation these degrees of freedom map onto a coupled theory of a vector-valued one-form gauge field and an ordinary $U(1)$ gauge field. We discuss the degrees of freedom in the corresponding g…
▽ More
We present a dual formulation of the Cosserat theory of elasticity. In this theory a local element of an elastic body is described in terms of local displacement and local orientation. Upon the duality transformation these degrees of freedom map onto a coupled theory of a vector-valued one-form gauge field and an ordinary $U(1)$ gauge field. We discuss the degrees of freedom in the corresponding gauge theories, the defect matter and coupling to the curved space.
△ Less
Submitted 17 March, 2020; v1 submitted 19 August, 2019;
originally announced August 2019.
-
Effective response theory for Floquet topological systems
Authors:
Paolo Glorioso,
Andrey Gromov,
Shinsei Ryu
Abstract:
We present an effective field theory approach to the topological response of Floquet systems with symmetry group $G$. This is achieved by introducing a background $G$ gauge field in the Schwinger-Keldysh formalism, which is suitable for far from equilibrium systems. We carry out this program for chiral topological Floquet systems (anomalous Floquet-Anderson insulators) in two spatial dimensions, a…
▽ More
We present an effective field theory approach to the topological response of Floquet systems with symmetry group $G$. This is achieved by introducing a background $G$ gauge field in the Schwinger-Keldysh formalism, which is suitable for far from equilibrium systems. We carry out this program for chiral topological Floquet systems (anomalous Floquet-Anderson insulators) in two spatial dimensions, and the group cohomology models of topological Floquet unitaries. These response actions serve as many-body topological invariants for topological Floquet unitaries. The effective action approach also leads us to propose novel topological response functions.
△ Less
Submitted 2 September, 2019; v1 submitted 8 August, 2019;
originally announced August 2019.
-
Numerical Simulations of Magnetized Astrophysical Jets and Comparison with Laboratory Laser Experiments
Authors:
V. S. Belyaev,
G. S. Bisnovatyi-Kogan,
A. I. Gromov,
B. V. Zagreev,
A. V. Lobanov,
A. P. Matafonov,
S. G. Moiseenko,
O. D. Toropina
Abstract:
The results of MHD numerical simulations of the formation and development of magnetized jets are presented. Similarity criteria for comparisons of the results of laboratory laser experiments and numerical simulations of astrophysical jets are discussed. The results of laboratory simulations of jets generated in experiments at the Neodim laser installation are presented.
The results of MHD numerical simulations of the formation and development of magnetized jets are presented. Similarity criteria for comparisons of the results of laboratory laser experiments and numerical simulations of astrophysical jets are discussed. The results of laboratory simulations of jets generated in experiments at the Neodim laser installation are presented.
△ Less
Submitted 11 March, 2019;
originally announced March 2019.
-
Towards classification of Fracton phases: the multipole algebra
Authors:
Andrey Gromov
Abstract:
We present an effective field theory approach to the Fracton phases. The approach is based the notion of a multipole algebra. It is an extension of space(-time) symmetries of a charge-conserving matter that includes global symmetries responsible for the conservation of various components of the multipole moments of the charge density. We explain how to construct field theories invariant under the…
▽ More
We present an effective field theory approach to the Fracton phases. The approach is based the notion of a multipole algebra. It is an extension of space(-time) symmetries of a charge-conserving matter that includes global symmetries responsible for the conservation of various components of the multipole moments of the charge density. We explain how to construct field theories invariant under the action of the algebra. These field theories generally break rotational invariance and exhibit anisotropic scaling. We further explain how to partially gauge the multipole algebra. Such gauging makes the symmetries responsible for the conservation of multipole moments local, while kee** rotation and translations symmetries global. It is shown that upon such gauging one finds the symmetric tensor gauge theories, as well as the generalized gauge theories discussed recently in the literature. The outcome of the gauging procedure depends on the choice of the multipole algebra. In particular, we show how to construct an effective theory for the $U(1)$ version of the Haah code based on the principles of symmetry and provide a two dimensional example with operators supported on a Sierpinski triangle. We show that upon condensation of charged excitations Fracton phases of both types as well as various SPTs emerge. Finally, the relation between the present approach and the formalism based on polynomials over finite fields is discussed.
△ Less
Submitted 12 December, 2018;
originally announced December 2018.
-
Geometric quench in the fractional quantum Hall effect: exact solution in quantum Hall matrix models and comparison with bimetric theory
Authors:
Matthew F. Lapa,
Andrey Gromov,
Taylor L. Hughes
Abstract:
We investigate the recently introduced geometric quench protocol for fractional quantum Hall (FQH) states within the framework of exactly solvable quantum Hall matrix models. In the geometric quench protocol a FQH state is subjected to a sudden change in the ambient geometry, which introduces anisotropy into the system. We formulate this quench in the matrix models and then we solve exactly for th…
▽ More
We investigate the recently introduced geometric quench protocol for fractional quantum Hall (FQH) states within the framework of exactly solvable quantum Hall matrix models. In the geometric quench protocol a FQH state is subjected to a sudden change in the ambient geometry, which introduces anisotropy into the system. We formulate this quench in the matrix models and then we solve exactly for the post-quench dynamics of the system and the quantum fidelity (Loschmidt echo) of the post-quench state. Next, we explain how to define a spin-2 collective variable $\hat{g}_{ab}(t)$ in the matrix models, and we show that for a weak quench (small anisotropy) the dynamics of $\hat{g}_{ab}(t)$ agrees with the dynamics of the intrinsic metric governed by the recently discussed bimetric theory of FQH states. We also find a modification of the bimetric theory such that the predictions of the modified bimetric theory agree with those of the matrix model for arbitrarily strong quenches. Finally, we introduce a class of higher-spin collective variables for the matrix model, which are related to generators of the $W_{\infty}$ algebra, and we show that the geometric quench induces nontrivial dynamics for these variables.
△ Less
Submitted 24 January, 2019; v1 submitted 17 September, 2018;
originally announced September 2018.
-
Geometric quench and nonequilibrium dynamics of fractional quantum Hall states
Authors:
Zhao Liu,
Andrey Gromov,
Zlatko Papić
Abstract:
We introduce a quench of the geometry of Landau level orbitals as a probe of nonequilibrium dynamics of fractional quantum Hall (FQH) states. We show that such geometric quenches induce coherent many-body dynamics of neutral degrees of freedom of FQH fluids. The simplest case of mass anisotropy quench can be experimentally implemented as a sudden tilt of the magnetic field, and the resulting dynam…
▽ More
We introduce a quench of the geometry of Landau level orbitals as a probe of nonequilibrium dynamics of fractional quantum Hall (FQH) states. We show that such geometric quenches induce coherent many-body dynamics of neutral degrees of freedom of FQH fluids. The simplest case of mass anisotropy quench can be experimentally implemented as a sudden tilt of the magnetic field, and the resulting dynamics reduces to the harmonic motion of the spin-$2$ "graviton" mode, i.e., the long wavelength limit of the Girvin-MacDonald-Platzman magnetoroton. We derive an analytical description of the graviton dynamics using the bimetric theory of FQH states, and find agreement with exact numerical simulations at short times. We show that certain types of geometric quenches excite higher-spin collective modes, thus establishing their existence in a microscopic model and motivating an extension of geometric theories of FQH states.
△ Less
Submitted 26 October, 2018; v1 submitted 28 February, 2018;
originally announced March 2018.
-
Measuring Electromagnetic and Gravitational Responses of Photonic Landau Levels
Authors:
Nathan Schine,
Michelle Chalupnik,
Tankut Can,
Andrey Gromov,
Jonathan Simon
Abstract:
The topology of an object describes global properties that are insensitive to local perturbations. Classic examples include string knots and the genus (number of handles) of a surface: no manipulation of a closed string short of cutting it changes its "knottedness"; and no deformation of a closed surface, short of puncturing it, changes how many handles it has. Topology has recently become an inte…
▽ More
The topology of an object describes global properties that are insensitive to local perturbations. Classic examples include string knots and the genus (number of handles) of a surface: no manipulation of a closed string short of cutting it changes its "knottedness"; and no deformation of a closed surface, short of puncturing it, changes how many handles it has. Topology has recently become an intense focus of condensed matter physics, where it arises in the context of the quantum Hall effect [1] and topological insulators [2]. In each case, topology is defined through invariants of the material's bulk [3-5], but experimentally measured through chiral/helical properties of the material's edges. In this work we measure topological invariants of a quantum Hall material through local response of the bulk: treating the material as a many-port circulator enables direct measurement of the Chern number as the spatial winding of the circulator phase; excess density accumulation near spatial curvature quantifies the curvature-analog of charge known as mean orbital spin, while the moment of inertia of this excess density reflects the chiral central charge. We observe that the topological invariants converge to their global values when probed over a few magnetic lengths lB, consistent with intuition that the bulk/edge distinction exists only for samples larger than a few lB. By performing these experiments in photonic Landau levels of a twisted resonator [6], we apply quantum-optics tools to topological matter. Combined with developments in Rydberg-mediated interactions between resonator photons [7], this work augurs an era of precision characterization of topological matter in strongly correlated fluids of light.
△ Less
Submitted 12 February, 2018;
originally announced February 2018.
-
Fractional quantum Hall systems near nematicity: bimetric theory, composite fermions, and Dirac brackets
Authors:
Dung Xuan Nguyen,
Andrey Gromov,
Dam Thanh Son
Abstract:
We perform a detailed comparison of the Dirac composite fermion and the recently proposed bimetric theory for a quantum Hall Jain states near half filling. By tuning the composite Fermi liquid to the vicinity of a nematic phase transition, we find that the two theories are equivalent to each other. We verify that the single mode approximation for the response functions and the static structure fac…
▽ More
We perform a detailed comparison of the Dirac composite fermion and the recently proposed bimetric theory for a quantum Hall Jain states near half filling. By tuning the composite Fermi liquid to the vicinity of a nematic phase transition, we find that the two theories are equivalent to each other. We verify that the single mode approximation for the response functions and the static structure factor becomes reliable near the phase transition. We show that the dispersion relation of the nematic mode near the phase transition can be obtained from the Dirac brackets between the components of the nematic order parameter. The dispersion is quadratic at low momenta and has a magnetoroton minimum at a finite momentum, which is not related to any nearby inhomogeneous phase.
△ Less
Submitted 21 December, 2017;
originally announced December 2017.
-
Chiral Topological Elasticity and Fracton Order
Authors:
Andrey Gromov
Abstract:
We analyze the "higher rank" gauge theories, that capture some of the phenomenology of the Fracton order. It is shown that these theories lose gauge invariance when arbitrarily weak and smooth curvature is introduced. We propose a resolution to this problem by introducing a theory invariant under area-preserving diffeomorphisms, which reduce to the "higher rank" gauge transformations upon lineariz…
▽ More
We analyze the "higher rank" gauge theories, that capture some of the phenomenology of the Fracton order. It is shown that these theories lose gauge invariance when arbitrarily weak and smooth curvature is introduced. We propose a resolution to this problem by introducing a theory invariant under area-preserving diffeomorphisms, which reduce to the "higher rank" gauge transformations upon linearization around a flat background. The proposed theory is geometric in nature and is interpreted as a theory of chiral topological elasticity. This theory exhibits some of the Fracton phenomenology. We explore the conservation laws, topological excitations, linear response, various kinematical constraints, and canonical structure of the theory. Finally, we emphasize that the very structure of Riemann-Cartan geometry, which we use to formulate the theory, encodes some of the Fracton phenomenology, suggesting that the Fracton order itself is geometric in nature.
△ Less
Submitted 18 February, 2019; v1 submitted 18 December, 2017;
originally announced December 2017.
-
Improved measurement of $^8$B solar neutrinos with 1.5 kt y of Borexino exposure
Authors:
The Borexino Collaboration,
M. Agostini,
K. Altenmüller,
S. Appel,
V. Atroshchenko,
Z. Bagdasarian,
D. Basilico,
G. Bellini,
J. Benziger,
D. Bick,
D. Bravo,
B. Caccianiga,
F. Calaprice,
A. Caminata,
P. Cavalcante,
A. Chepurnov,
D. D'Angelo,
S. Davini,
A. Derbin,
A. Di Giacinto,
V. Di Marcello,
X. F. Ding,
A. Di Ludovico,
L. Di Noto,
I. Drachnev
, et al. (73 additional authors not shown)
Abstract:
We report on an improved measurement of the $^8$B solar neutrino interaction rate with the Borexino experiment at the Laboratori Nazionali del Gran Sasso. Neutrinos are detected via their elastic scattering on electrons in a large volume of liquid scintillator. The measured rate of scattered electrons above 3 MeV of energy is…
▽ More
We report on an improved measurement of the $^8$B solar neutrino interaction rate with the Borexino experiment at the Laboratori Nazionali del Gran Sasso. Neutrinos are detected via their elastic scattering on electrons in a large volume of liquid scintillator. The measured rate of scattered electrons above 3 MeV of energy is $0.223\substack{+0.015 \\ -0.016}\,(stat)\,\substack{+0.006 \\ -0.006}\,(syst)$ cpd/100 t, which corresponds to an observed solar neutrino flux assuming no neutrino flavor conversion of $Φ\substack{\rm ES \\ ^8\rm B}=2.57\substack{+0.17 \\ -0.18}(stat)\substack{+0.07\\ -0.07}(syst)\times$10$^6$ cm$^{-2}\,$s$^{-1}$. This measurement exploits the active volume of the detector in almost its entirety for the first time, and takes advantage of a reduced radioactive background following the 2011 scintillator purification campaign and of novel analysis tools providing a more precise modeling of the background. Additionally, we set a new limit on the interaction rate of solar $hep$ neutrinos, searched via their elastic scattering on electrons as well as their neutral current-mediated inelastic scattering on carbon, $^{12}$C($ν,ν'$)$^{12}$C* ($E_γ$= 15.1 MeV).
△ Less
Submitted 6 March, 2020; v1 submitted 3 September, 2017;
originally announced September 2017.
-
Transport signatures of Hall viscosity
Authors:
Luca V. Delacretaz,
Andrey Gromov
Abstract:
Hall viscosity is a non-dissipative response function describing momentum transport in two-dimensional systems with broken parity. It is quantized in the quantum Hall regime, and contains information about the topological order of the quantum Hall state. Hall viscosity can distinguish different quantum Hall states with identical Hall conductances, but different topological order. To date, an exper…
▽ More
Hall viscosity is a non-dissipative response function describing momentum transport in two-dimensional systems with broken parity. It is quantized in the quantum Hall regime, and contains information about the topological order of the quantum Hall state. Hall viscosity can distinguish different quantum Hall states with identical Hall conductances, but different topological order. To date, an experimentally accessible signature of Hall viscosity is lacking. We exploit the fact that Hall viscosity contributes to charge transport at finite wavelengths, and can therefore be extracted from non-local resistance measurements in inhomogeneous charge flows. We explain how to determine the Hall viscosity from such a transport experiment. In particular, we show that the profile of the electrochemical potential close to contacts where current is injected is sensitive to the value of the Hall viscosity.
△ Less
Submitted 21 September, 2017; v1 submitted 12 June, 2017;
originally announced June 2017.
-
Bimetric Theory of Fractional Quantum Hall States
Authors:
Andrey Gromov,
Dam Thanh Son
Abstract:
We present a bimetric low-energy effective theory of fractional quantum Hall (FQH) states that describes the topological properties and a gapped collective excitation, known as Girvin-Macdonald-Platzman (GMP) mode. The theory consist of a topological Chern-Simons action, coupled to a symmetric rank two tensor, and an action à la bimetric gravity, describing the gapped dynamics of the spin-$2$ GMP…
▽ More
We present a bimetric low-energy effective theory of fractional quantum Hall (FQH) states that describes the topological properties and a gapped collective excitation, known as Girvin-Macdonald-Platzman (GMP) mode. The theory consist of a topological Chern-Simons action, coupled to a symmetric rank two tensor, and an action à la bimetric gravity, describing the gapped dynamics of the spin-$2$ GMP mode. The theory is formulated in curved ambient space and is spatially covariant, which allows to restrict the form of the effective action and the values of phenomenological coefficients. Using the bimetric theory we calculate the projected static structure factor up to the $k^6$ order in the momentum expansion. To provide further support for the theory, we derive the long wave limit of the GMP algebra, the dispersion relation of the GMP mode, and the Hall viscosity of FQH states. We also comment on the possible applications to fractional Chern insulators, where closely related structures arise. Finally, it is shown that the familiar FQH observables acquire a curious geometric interpretation within the bimetric formalism.
△ Less
Submitted 14 November, 2017; v1 submitted 18 May, 2017;
originally announced May 2017.
-
Parameters of Three Selected Model Galactic Potentials Based on the Velocities of Objects at Distances up to 200 kpc
Authors:
V. V. Bobylev,
A. T. Bajkova,
A. O. Gromov
Abstract:
This paper is a continuation of our recent paper devoted to refining the parameters of three component (bulge, disk, halo) axisymmetric model Galactic gravitational potentials differing by the expression for the dark matter halo using the velocities of distant objects. In all models the bulge and disk potentials are described by the Miyamoto-Nagai expressions. In our previous paper we used the All…
▽ More
This paper is a continuation of our recent paper devoted to refining the parameters of three component (bulge, disk, halo) axisymmetric model Galactic gravitational potentials differing by the expression for the dark matter halo using the velocities of distant objects. In all models the bulge and disk potentials are described by the Miyamoto-Nagai expressions. In our previous paper we used the Allen-Santill'an (I), Wilkinson--Evans (II), and Navarro-Frenk-White (III) models to describe the halo. In this paper we use a spherical logarithmic Binney potential (model IV), a Plummer sphere (model V), and a Hernquist potential (model VI) to describe the halo. A set of present-day observational data in the range of Galactocentric distances R from 0 to 200 kpc is used to refine the parameters of the listed models, which are employed most commonly at present. The model rotation curves are fitted to the observed velocities by taking into account the constraints on the local matter density and the vertical force . Model VI looks best among the three models considered here from the viewpoint of the achieved accuracy of fitting the model rotation curves to the measurements. This model is close to the Navarro-Frenk-White model III refined and considered best in our previous paper, which is shown using the integration of the orbits of two globular clusters, Lynga 7 and NGC 5053, as an example.
△ Less
Submitted 4 March, 2017;
originally announced March 2017.
-
Investigating anisotropic quantum Hall states with bi-metric geometry
Authors:
Andrey Gromov,
Scott D. Geraedts,
Barry Bradlyn
Abstract:
We construct a low energy effective theory of anisotropic fractional quantum Hall (FQH) states. We develop a formalism similar to that used in the bi-metric approach to massive gravity, and apply it to describe abelian anisotropic FQH states in the presence of external electromagnetic and geometric backgrounds. We derive a relationship between the shift, the Hall viscosity, and a new quantized cou…
▽ More
We construct a low energy effective theory of anisotropic fractional quantum Hall (FQH) states. We develop a formalism similar to that used in the bi-metric approach to massive gravity, and apply it to describe abelian anisotropic FQH states in the presence of external electromagnetic and geometric backgrounds. We derive a relationship between the shift, the Hall viscosity, and a new quantized coupling to anisotropy, which we term "anisospin". We verify this relationship by numerically computing the Hall viscosity for a variety of anisotropic quantum Hall states using the density matrix renormalization group (DMRG). Finally, we apply these techniques to the problem of nematic order and clarify certain disagreements that exist in the literature about the meaning of the coefficient of the Berry phase term in the nematic effective action.
△ Less
Submitted 27 November, 2017; v1 submitted 3 March, 2017;
originally announced March 2017.
-
Anyonic Chains, Topological Defects, and Conformal Field Theory
Authors:
Matthew Buican,
Andrey Gromov
Abstract:
Motivated by the three-dimensional topological field theory / two-dimensional conformal field theory (CFT) correspondence, we study a broad class of one-dimensional quantum mechanical models, known as anyonic chains, that can give rise to an enormously rich (and largely unexplored) space of two-dimensional critical theories in the thermodynamic limit. One remarkable feature of these systems is the…
▽ More
Motivated by the three-dimensional topological field theory / two-dimensional conformal field theory (CFT) correspondence, we study a broad class of one-dimensional quantum mechanical models, known as anyonic chains, that can give rise to an enormously rich (and largely unexplored) space of two-dimensional critical theories in the thermodynamic limit. One remarkable feature of these systems is the appearance of non-local microscopic "topological symmetries" that descend to topological defects of the resulting CFTs. We derive various model-independent properties of these theories and of this topological symmetry / topological defect correspondence. For example, by studying precursors of certain twist and defect fields directly in the anyonic chains, we argue that (under mild assumptions) the two-dimensional CFTs correspond to particular modular invariants with respect to their maximal chiral algebras and that the topological defects descending from topological symmetries commute with these maximal chiral algebras. Using this map, we apply properties of defect Hilbert spaces to show how topological symmetries give a handle on the set of allowed relevant deformations of these theories. Throughout, we give a unified perspective that treats the constraints from discrete symmetries on the same footing as the constraints from topological ones.
△ Less
Submitted 20 February, 2017; v1 submitted 10 January, 2017;
originally announced January 2017.
-
Particle-Hole Duality in the Lowest Landau Level
Authors:
Dung Xuan Nguyen,
Tankut Can,
Andrey Gromov
Abstract:
We derive a number of exact relations between response functions of holomorphic, chiral fractional quantum Hall states and their particle-hole (PH) conjugates. These exact relations allow one to calculate the Hall conductivity, Hall viscosity, various Berry phases, and the static structure factor of PH-conjugate states from the corresponding properties of the original states. These relations estab…
▽ More
We derive a number of exact relations between response functions of holomorphic, chiral fractional quantum Hall states and their particle-hole (PH) conjugates. These exact relations allow one to calculate the Hall conductivity, Hall viscosity, various Berry phases, and the static structure factor of PH-conjugate states from the corresponding properties of the original states. These relations establish a precise duality between chiral quantum Hall states and their PH-conjugates. The key ingredient in the proof of the relations is a generalization of Girvin's construction of PH-conjugate states to inhomogeneous magnetic field and curvature. Finally, we make several non-trivial checks of the relations, including for the Jain states and their PH-conjugates.
△ Less
Submitted 22 December, 2016;
originally announced December 2016.