-
Multi-block linearized alternating direction method for sparse fused Lasso modeling problems
Authors:
Xiaofei Wu,
Rongmei Liang,
Zhimin Zhang,
Zhenyu Cui
Abstract:
In many statistical modeling problems, such as classification and regression, it is common to encounter sparse and blocky coefficients. Sparse fused Lasso is specifically designed to recover these sparse and blocky structured features, especially in cases where the design matrix has ultrahigh dimensions, meaning that the number of features significantly surpasses the number of samples. Quantile lo…
▽ More
In many statistical modeling problems, such as classification and regression, it is common to encounter sparse and blocky coefficients. Sparse fused Lasso is specifically designed to recover these sparse and blocky structured features, especially in cases where the design matrix has ultrahigh dimensions, meaning that the number of features significantly surpasses the number of samples. Quantile loss is a well-known robust loss function that is widely used in statistical modeling. In this paper, we propose a new sparse fused lasso classification model, and develop a unified multi-block linearized alternating direction method of multipliers algorithm that effectively selects sparse and blocky features for regression and classification. Our algorithm has been proven to converge with a derived linear convergence rate. Additionally, our algorithm has a significant advantage over existing methods for solving ultrahigh dimensional sparse fused Lasso regression and classification models due to its lower time complexity. Note that the algorithm can be easily extended to solve various existing fused Lasso models. Finally, we present numerical results for several synthetic and real-world examples, which demonstrate the robustness, scalability, and accuracy of the proposed classification model and algorithm
△ Less
Submitted 29 May, 2024; v1 submitted 18 November, 2023;
originally announced November 2023.
-
Algorithmic stability implies training-conditional coverage for distribution-free prediction methods
Authors:
Ruiting Liang,
Rina Foygel Barber
Abstract:
In a supervised learning problem, given a predicted value that is the output of some trained model, how can we quantify our uncertainty around this prediction? Distribution-free predictive inference aims to construct prediction intervals around this output, with valid coverage that does not rely on assumptions on the distribution of the data or the nature of the model training algorithm. Existing…
▽ More
In a supervised learning problem, given a predicted value that is the output of some trained model, how can we quantify our uncertainty around this prediction? Distribution-free predictive inference aims to construct prediction intervals around this output, with valid coverage that does not rely on assumptions on the distribution of the data or the nature of the model training algorithm. Existing methods in this area, including conformal prediction and jackknife+, offer theoretical guarantees that hold marginally (i.e., on average over a draw of training and test data). In contrast, training-conditional coverage is a stronger notion of validity that ensures predictive coverage of the test point for most draws of the training data, and is thus a more desirable property in practice. Training-conditional coverage was shown by Vovk [2012] to hold for the split conformal method, but recent work by Bian and Barber [2023] proves that such validity guarantees are not possible for the full conformal and jackknife+ methods without further assumptions. In this paper, we show that an assumption of algorithmic stability ensures that the training-conditional coverage property holds for the full conformal and jackknife+ methods.
△ Less
Submitted 25 June, 2024; v1 submitted 7 November, 2023;
originally announced November 2023.
-
Optimal divergence rate of the focusing Gibbs measure
Authors:
Guopeng Li,
Rui Liang,
Yuzhao Wang
Abstract:
We study the focusing Gibbs measure with critical/supercritical potentials. In particular, we prove asymptotic formulae for the frequency approximation of the partition function, which captures the optimal divergence rate of the partition function as the frequency truncation is removed.
We study the focusing Gibbs measure with critical/supercritical potentials. In particular, we prove asymptotic formulae for the frequency approximation of the partition function, which captures the optimal divergence rate of the partition function as the frequency truncation is removed.
△ Less
Submitted 12 October, 2023;
originally announced October 2023.
-
Gibbs Dynamics for the Weakly Dispersive Nonlinear Schrödinger Equations
Authors:
Rui Liang,
Yuzhao Wang
Abstract:
We consider the Cauchy problem for the one-dimensional periodic cubic nonlinear fractional Schrödinger equation (FNLS) with initial data distributed via Gibbs measure. We construct global strong solutions with the flow property to FNLS on the support of the Gibbs measure in the full dispersive range, thus resolving a question proposed by Sun-Tzvetkov (2021). As a byproduct, we prove the invariance…
▽ More
We consider the Cauchy problem for the one-dimensional periodic cubic nonlinear fractional Schrödinger equation (FNLS) with initial data distributed via Gibbs measure. We construct global strong solutions with the flow property to FNLS on the support of the Gibbs measure in the full dispersive range, thus resolving a question proposed by Sun-Tzvetkov (2021). As a byproduct, we prove the invariance of the Gibbs measure and almost sure global well-posedness for FNLS.
△ Less
Submitted 13 June, 2023;
originally announced June 2023.
-
Unbounded Chaotic Weighted Pseudo-shift
Authors:
Ruxi Liang,
Pengyu Qin,
Yonglu Shu
Abstract:
In this paper, based on the work of Vijay K. Srivastava and Harish Chandra, we give a characterization of the unbounded hypercyclic weighted pseudo-shift operator $wC_{\varphi}$ on $\ell^p$ or $c_0$. Moreover we use the hypercyclicity criterion of Bès, Chan, and Seubert to give a necessary and sufficient condition in order that $wC_{\varphi}$ be a chaotic operator.
In this paper, based on the work of Vijay K. Srivastava and Harish Chandra, we give a characterization of the unbounded hypercyclic weighted pseudo-shift operator $wC_{\varphi}$ on $\ell^p$ or $c_0$. Moreover we use the hypercyclicity criterion of Bès, Chan, and Seubert to give a necessary and sufficient condition in order that $wC_{\varphi}$ be a chaotic operator.
△ Less
Submitted 10 May, 2023; v1 submitted 12 March, 2023;
originally announced March 2023.
-
Semi-overlap Functions and Novel Fuzzy Reasoning Algorithms with Applications
Authors:
Xiaohong Zhang,
Mei Wang,
Benjamin Bedregal,
Mengyuan Li,
Rong Liang
Abstract:
It is worth noticing that a fuzzy conjunction and its corresponding fuzzy implication can form a residual pair if and only if it is left-continuous. In order to get a more general result related on residual implications that induced by aggregation functions, we relax the definition of general overlap functions, more precisely, removing its right-continuous, and then introduce a new kind of aggrega…
▽ More
It is worth noticing that a fuzzy conjunction and its corresponding fuzzy implication can form a residual pair if and only if it is left-continuous. In order to get a more general result related on residual implications that induced by aggregation functions, we relax the definition of general overlap functions, more precisely, removing its right-continuous, and then introduce a new kind of aggregation functions, which called semi-overlap functions. Subsequently, we study some of their related algebraic properties and investigate their corresponding residual implications. Moreover, serval scholars have provided kinds of methods for fuzzy modus ponens (FMP,for short) problems so far, such as Zadeh's compositional rule of inference (CRI, for short), Wang's triple I method (TIM, for short) and the quintuple implication principle (QIP, for short). Compared with CRI and TIM methods, QIP method has some advantages in solving FMP problems, in this paper, we further consider the QIP method for FMP problems and prove that it satisfies the reducibility of multiple-rules fuzzy reasoning. Finally, we propose a new classification algorithm that based on semi-overlap functions and QIP method, which called SO5I-FRC algorithm. Through the comparative tests, the average accuracy of SO5I-FRC algorithm is higher than FARC-HD algorithm. The experimental results indicate that semi-overlap functions and QIP method have certain advantages and a wide range of applications in classification problems.
△ Less
Submitted 1 March, 2022;
originally announced March 2022.
-
Gibbs measure for the focusing fractional NLS on the torus
Authors:
Rui Liang,
Yuzhao Wang
Abstract:
We study the construction of the Gibbs measures for the {\it focusing} mass-critical fractional nonlinear Schrödinger equation on the multi-dimensional torus. We identify the sharp mass threshold for normalizability and non-normalizability of the focusing Gibbs measures, which generalizes the influential works of Lebowitz-Rose-Speer (1988), Bourgain (1994), and Oh-Sosoe-Tolomeo (2021) on the one-d…
▽ More
We study the construction of the Gibbs measures for the {\it focusing} mass-critical fractional nonlinear Schrödinger equation on the multi-dimensional torus. We identify the sharp mass threshold for normalizability and non-normalizability of the focusing Gibbs measures, which generalizes the influential works of Lebowitz-Rose-Speer (1988), Bourgain (1994), and Oh-Sosoe-Tolomeo (2021) on the one-dimensional nonlinear Schrödinger equations. To this purpose, we establish an almost sharp fractional Gagliardo-Nirenberg-Sobolev inequality on the torus, which is of independent interest.
△ Less
Submitted 17 April, 2023; v1 submitted 12 September, 2021;
originally announced September 2021.
-
Detecting Carbon Nanotube Orientation with Topological Data Analysis of Scanning Electron Micrographs
Authors:
Liyu Dong,
Haibin Hang,
** Gyu Park,
Washington Mio,
Richard Liang
Abstract:
As the aerospace industry becomes increasingly demanding for stronger lightweight materials, the ultra-strong carbon nanotube (CNT) composites with highly aligned CNT network structures could be the answer. In this work, a novel methodology applying topological data analysis (TDA) to the scanning electron microscope (SEM) images was developed to detect CNT orientation. The CNT bundle extensions in…
▽ More
As the aerospace industry becomes increasingly demanding for stronger lightweight materials, the ultra-strong carbon nanotube (CNT) composites with highly aligned CNT network structures could be the answer. In this work, a novel methodology applying topological data analysis (TDA) to the scanning electron microscope (SEM) images was developed to detect CNT orientation. The CNT bundle extensions in certain directions were summarized algebraically and expressed as visible barcodes. The barcodes were then calculated and converted into the total spread function $V(X,θ)$, from which the alignment fraction and the preferred direction could be determined. For validation purposes, the random CNT sheets were mechanically stretched at various strain ratios ranging from $0-40\%$, and quantitative TDA analysis was conducted based on the SEM images taken at random positions. The results showed high consistency ($R^2=0.975$) compared to the Herman's orientation factors derived from the polarized Raman spectroscopy and wide-angle X-ray scattering analysis. Additionally, the TDA method presented great robustness with varying SEM acceleration voltages and magnifications, which might alter the scope in alignment detection. With potential applications in nanofiber systems, this study offers a rapid and simple way to quantify CNT alignment, which plays a crucial role in transferring the CNT properties into engineering products.
△ Less
Submitted 9 August, 2021;
originally announced August 2021.
-
Packing entropy for fixed-point free flows
Authors:
Ruiming Liang,
Haoyi Lei
Abstract:
Let $(X,φ)$ be a compact flow without fixed points. We define the packing topological entropy $h_{\mathrm{top}}^P(φ,K)$ on subsets of $X$ through considering all the possible reparametrizations of time. For fixed-point free flows, we prove the following result: for any non-empty compact subset $K$ of $X$,…
▽ More
Let $(X,φ)$ be a compact flow without fixed points. We define the packing topological entropy $h_{\mathrm{top}}^P(φ,K)$ on subsets of $X$ through considering all the possible reparametrizations of time. For fixed-point free flows, we prove the following result: for any non-empty compact subset $K$ of $X$, $$h_{\mathrm{top}}^P(φ,K)=\sup\{\overline{h}_μ(φ):μ(K)=1,μ\text{ is a Borel probability measure on} X\},$$ where $\overline{h}_μ(φ)$ denotes the upper local entropy for a Borel probability measure $μ$ on $X$.
△ Less
Submitted 20 February, 2021;
originally announced February 2021.
-
Splitting and successively solving augmented Lagrangian method for optimization with semicontinuous variables and cardinality constraint
Authors:
Yanqin Bai,
Renli Liang,
Zhouwang Yang
Abstract:
We propose a new splitting and successively solving augmented Lagrangian (SSAL) method for solving an optimization problem with both semicontinuous variables and a cardinality constraint. This optimization problem arises in several contexts such as the portfolio selection problem, the compressed sensing problem and the unit commitment problem, etc. The problem is in general NP-hard. We derive an o…
▽ More
We propose a new splitting and successively solving augmented Lagrangian (SSAL) method for solving an optimization problem with both semicontinuous variables and a cardinality constraint. This optimization problem arises in several contexts such as the portfolio selection problem, the compressed sensing problem and the unit commitment problem, etc. The problem is in general NP-hard. We derive an optimality condition for this optimization problem, under some suitable assumptions. By introducing an auxiliary variable and using an augmented Lagrangian function, the constraints are decomposed into two parts. By fixing particular variables, the optimization problem is split into two subproblems, which are solved alternatively. Furthermore, we prove the convergence of SSAL, under some suitable assumptions. Finally, we implement our method for the portfolio selection problem and the compressed sensing problem, respectively. Real world data and simulation study show that SSAL outperforms the well-known CPLEX 12.6 and the penalty decomposition (PD) method, while enjoying a similar cardinality of decision variable. For example, the numerical results are even nearly 200 times faster than that of CPLEX 12.6 for the portfolio selection problem, and more than 40 times faster than PD method for the compressed sensing problem, respectively. In particularly, SSAL is powerful when the size of problem increase largely.
△ Less
Submitted 15 June, 2015;
originally announced June 2015.
