Minimax Concave Penalty

The minimax concave penalty (MCP) is a non-convex regularization technique used primarily to enhance the sparsity and accuracy of statistical models, particularly in high-dimensional settings. Current research focuses on applying MCP within optimization frameworks like ADMM and proximal gradient methods, often in conjunction with smoothing techniques, to address challenges posed by its non-convexity, especially in distributed or federated learning scenarios for quantile regression. This work aims to improve the efficiency and convergence properties of algorithms employing MCP, leading to more robust and accurate model estimation in various applications, including those involving large-scale datasets from IoT devices and tensor recovery problems. The resulting improvements in model accuracy and computational efficiency have significant implications for diverse fields like signal processing, machine learning, and data analysis.

Papers

August 10, 2024

Federated Smoothing Proximal Gradient for Quantile Regression with Non-Convex Penalties
Reza Mirzaeifard, Diyako Ghaderyan, Stefan Werner
Proximal Gradient Federated Stochastic Non Convex Penalty Minimax Concave Penalty Sparse Penalty

August 2, 2024

Decentralized Smoothing ADMM for Quantile Regression with Non-Convex Sparse Penalties
Reza Mirzaeifard, Diyako Ghaderyan, Stefan Werner
ADMM Algorithm Decentralized Kernel Ridge Regression Non Convex Penalty Minimax Concave Penalty Sparse Penalty

September 4, 2023

Smoothing ADMM for Sparse-Penalized Quantile Regression with Non-Convex Penalties
Reza Mirzaeifard, Naveen K. D. Venkategowda, Vinay Chakravarthi Gogineni, Stefan Werner
Alternating Direction Method of Multiplier Non Convex Penalty Minimax Concave Penalty Sparse Penalty

May 31, 2023

ROSARL: Reward-Only Safe Reinforcement Learning
Geraud Nangue Tasse, Tamlin Love, Mark Nemecek, Steven James, Benjamin Rosman
Reinforcement Learning High Dimensional Optimal Policy Safe Reinforcement Learning Model Free Algorithm Minimax Concave Penalty

June 25, 2022

Tensor Recovery Based on A Novel Non-convex Function Minimax Logarithmic Concave Penalty Function
Hongbing Zhang, Xinyi Liu, Chang Liu, Hongtao Fan, Yajing Li, Xinyun Zhu
Convex Relaxation Tensor Recovery Minimax Concave Penalty

January 30, 2022

Low-Rank Tensor Completion Based on Bivariate Equivalent Minimax-Concave Penalty
Hongbing Zhang, Xinyi Liu, Hongtao Fan, Yajing Li, Yinlin Ye
Gradient Norm Low Rank Tensor Completion Minimax Concave Penalty

Minimax Concave Penalty

Papers

Federated Smoothing Proximal Gradient for Quantile Regression with Non-Convex Penalties

Decentralized Smoothing ADMM for Quantile Regression with Non-Convex Sparse Penalties

Smoothing ADMM for Sparse-Penalized Quantile Regression with Non-Convex Penalties

ROSARL: Reward-Only Safe Reinforcement Learning

Tensor Recovery Based on A Novel Non-convex Function Minimax Logarithmic Concave Penalty Function

Low-Rank Tensor Completion Based on Bivariate Equivalent Minimax-Concave Penalty