Variance Reduction

Variance reduction techniques aim to improve the efficiency and stability of various machine learning and optimization algorithms by decreasing the variability in estimates of gradients or other key quantities. Current research focuses on applying these techniques to diverse areas, including meta-learning, reinforcement learning, federated learning, and Monte Carlo methods, often employing algorithms like stochastic variance reduced gradient (SVRG), control variates, and importance sampling. These advancements lead to faster convergence, improved sample efficiency, and more robust performance in various applications, ultimately impacting the scalability and reliability of machine learning models and simulations.

Papers

February 3, 2023

Rethinking Semi-Supervised Medical Image Segmentation: A Variance-Reduction Perspective
Chenyu You, Weicheng Dai, Yifei Min, Fenglin Liu, David A. Clifton, S Kevin Zhou, Lawrence Hamilton Staib, James S Duncan
Contrastive Learning Medical Image Segmentation Semi Supervised Semi Supervised Medical Image Segmentation Variance Reduction Voxel Based Semantic Segmentation

January 30, 2023

SoftTreeMax: Exponential Variance Reduction in Policy Gradient via Tree Search
Gal Dalal, Assaf Hallak, Gugan Thoppe, Shie Mannor, Gal Chechik
Policy Gradient Softmax Function Variance Reduction Tree Search Gradient Variance

January 2, 2023

Stochastic Variable Metric Proximal Gradient with variance reduction for non-convex composite optimization
Gersende Fort, Eric Moulines
Variance Reduction Proximal Point Composite Optimization Stochastic Proximal Gradient

December 27, 2022

Variance Reduction for Score Functions Using Optimal Baselines
Ronan Keane, H. Oliver Gao
Value Function Variance Reduction Gradient Estimator Score Function Optimal Baseline

December 9, 2022

Cyclic Block Coordinate Descent With Variance Reduction for Composite Nonconvex Optimization
Xufeng Cai, Chaobing Song, Stephen J. Wright, Jelena Diakonikolas
Gradient Norm Variance Reduction Nonconvex Optimization Block Coordinate Nonsmooth Composite Optimization

December 5, 2022

On the effectiveness of partial variance reduction in federated learning with heterogeneous data
Bo Li, Mikkel N. Schmidt, Tommy S. Alstrøm, Sebastian U. Stich
LeArning Abstract Data Heterogeneity Heterogeneous Data Variance Reduction FedAvg Algorithm Gradient Update Client Drift

November 29, 2022

Approximating Martingale Process for Variance Reduction in Deep Reinforcement Learning with Large State Space
Charlie Ruan
Reinforcement Learning Deep Reinforcement Learning Proximal Policy Optimization State Space Variance Reduction

November 15, 2022

An Improved Analysis of (Variance-Reduced) Policy Gradient and Natural Policy Gradient Methods
Yanli Liu, Kaiqing Zhang, Tamer Başar, Wotao Yin
Policy Gradient Variance Reduction Natural Policy Gradient Improved Analysis Policy Parametrization Variance Aware

November 3, 2022

Adaptive Stochastic Variance Reduction for Non-convex Finite-Sum Minimization
Ali Kavis, Stratis Skoulakis, Kimon Antonakopoulos, Leello Tadesse Dadi, Volkan Cevher
Gradient Norm Variance Reduction Finite Sum Minimization Stochastic Path Adaptive Variance

October 12, 2022

SARAH-based Variance-reduced Algorithm for Stochastic Finite-sum Cocoercive Variational Inequalities
Aleksandr Beznosikov, Alexander Gasnikov
Variance Reduction Variational Inequality Monotone Function Stochastic Recursive Gradient Stochastic Approach

September 29, 2022

META-STORM: Generalized Fully-Adaptive Variance Reduced SGD for Unbounded Functions
Zijian Liu, Ta Duy Nguyen, Thien Hang Nguyen, Alina Ene, Huy L. Nguyen
Adaptive Importance Stochastic Gradient Variance Reduction Non Convex Optimization Non Convex Stochastic Optimization Variance Reduced Unbounded Gradient Medium STORM

September 16, 2022

Reducing Variance in Temporal-Difference Value Estimation via Ensemble of Deep Networks
Litian Liang, Yaosheng Xu, Stephen McAleer, Dailin Hu, Alexander Ihler, Pieter Abbeel, Roy Fox
Deep Network Diverse Ensemble Variance Reduction Temporal Difference Learning Temporal Difference Variance Estimation Atari Benchmark

August 25, 2022

Variance Reduction based Experience Replay for Policy Optimization
Hua Zheng, Wei Xie, M. Ben Feng
Reinforcement Learning Policy OpTimization Policy Optimization Variance Reduction Experience Replay

August 10, 2022

Fast Heterogeneous Federated Learning with Hybrid Client Selection
Guangyuan Shen, Dehong Gao, Duanxiao Song, Libin Yang, Xukai Zhou, Shirui Pan, Wei Lou, Fang Zhou
Variance Reduction Heterogeneous Federated Learning Client Selection Cluster Specific Random Effect

July 28, 2022

FedVARP: Tackling the Variance Due to Partial Client Participation in Federated Learning
Divyansh Jhunjhunwala, Pranay Sharma, Aushim Nagarkatti, Gauri Joshi
Variance Reduction Local Optimization Client Participation Data Variance

July 8, 2022

Tackling Data Heterogeneity: A New Unified Framework for Decentralized SGD with Sample-induced Topology
Yan Huang, Ying Sun, Zehan Zhu, Changzhi Yan, Jinming Xu
Unified Framework Empirical Risk Minimization Data Heterogeneity Convergence Analysis Variance Reduction Topological Metric Decentralized SGD Stochastic Gradient Based

June 14, 2022

June 6, 2022

Stochastic Variance-Reduced Newton: Accelerating Finite-Sum Minimization with Large Batches
Michał Dereziński
Second Order Stochastic Way Variance Reduction Subsampling Method Large Batch Newton Method Finite Sum Minimization Variance Reduced Stochastic

June 1, 2022

Variance Reduction is an Antidote to Byzantines: Better Rates, Weaker Assumptions and Communication Compression as a Cherry on the Top
Eduard Gorbunov, Samuel Horváth, Peter Richtárik, Gauthier Gidel
Loss Function Variance Reduction Byzantine Robust Fast Rate Communication Compression Byzantine Seal Byzantine Robust Method

Variance Reduction

Papers

Rethinking Semi-Supervised Medical Image Segmentation: A Variance-Reduction Perspective

SoftTreeMax: Exponential Variance Reduction in Policy Gradient via Tree Search

Stochastic Variable Metric Proximal Gradient with variance reduction for non-convex composite optimization

Variance Reduction for Score Functions Using Optimal Baselines

Cyclic Block Coordinate Descent With Variance Reduction for Composite Nonconvex Optimization

On the effectiveness of partial variance reduction in federated learning with heterogeneous data

Approximating Martingale Process for Variance Reduction in Deep Reinforcement Learning with Large State Space

An Improved Analysis of (Variance-Reduced) Policy Gradient and Natural Policy Gradient Methods

Adaptive Stochastic Variance Reduction for Non-convex Finite-Sum Minimization

SARAH-based Variance-reduced Algorithm for Stochastic Finite-sum Cocoercive Variational Inequalities

META-STORM: Generalized Fully-Adaptive Variance Reduced SGD for Unbounded Functions

Reducing Variance in Temporal-Difference Value Estimation via Ensemble of Deep Networks

Variance Reduction based Experience Replay for Policy Optimization

Fast Heterogeneous Federated Learning with Hybrid Client Selection

FedVARP: Tackling the Variance Due to Partial Client Participation in Federated Learning

Tackling Data Heterogeneity: A New Unified Framework for Decentralized SGD with Sample-induced Topology

Tearing Apart NOTEARS: Controlling the Graph Prediction via Variance Manipulation

Variance Reduction for Policy-Gradient Methods via Empirical Variance Minimization

Stochastic Variance-Reduced Newton: Accelerating Finite-Sum Minimization with Large Batches

Variance Reduction is an Antidote to Byzantines: Better Rates, Weaker Assumptions and Communication Compression as a Cherry on the Top