DFWLayer: Differentiable Frank-Wolfe Optimization Layer

Zixuan Liu, Liu Liu, Xueqian Wang, Peilin Zhao
8/21/2023
cs.LGcs.AI

No official code found. Best implementation: Panda-Shawn/DFWLayer (2 stars, MIT, last commit 2024-07-25).

Abstract

Differentiable optimization has received a significant amount of attention due to its foundational role in the domain of machine learning based on neural networks. This paper proposes a differentiable layer, named Differentiable Frank-Wolfe Layer (DFWLayer), by rolling out the Frank-Wolfe method, a well-known optimization algorithm which can solve constrained optimization problems without projections and Hessian matrix computations, thus leading to an efficient way of dealing with large-scale convex optimization problems with norm constraints. Experimental results demonstrate that the DFWLayer not only attains competitive accuracy in solutions and gradients but also consistently adheres to constraints.

View on arXivView PDF

Code Implementations(1)

Panda-Shawn/DFWLayerCommunity100%
MIT

DFWLayer: Differentiable Frank-Wolfe Optimization Layer

20PythonAug 25, 20232 years agoMIT

Similar Papers

Karush-Kuhn-Tucker Condition-Trained Neural Networks (KKT Nets)

88% match2024

SANIA: Polyak-type Optimization Framework Leads to Scale Invariant Stochastic Algorithms

87% match2023

Differentiable Convex Optimization Layers in Neural Architectures: Foundations and Perspectives

86% match2024

Multi-Objective Optimization for Sparse Deep Multi-Task Learning

85% match1 repo162023

Non-Convex Optimizations for Machine Learning with Theoretical Guarantee: Robust Matrix Completion and Neural Network Learning

85% match2023

Review Non-convex Optimization Method for Machine Learning

85% match2024

Cite this paper

@article{liu2023dfwlayer,
  title  = {DFWLayer: Differentiable Frank-Wolfe Optimization Layer},
  author = {Zixuan Liu and Liu Liu and Xueqian Wang and Peilin Zhao},
  year   = {2023},
  eprint = {2308.10806},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2308.10806v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2308.10806.svg)](https://codeofpaper.com/paper/2308.10806)

Discussion