Picture for Dixian Zhu

Dixian Zhu

Function Aligned Regression: A Method Explicitly Learns Functional Derivatives from Data

Add code
Feb 08, 2024
Viaarxiv icon

Non-Smooth Weakly-Convex Finite-sum Coupled Compositional Optimization

Add code
Oct 05, 2023
Viaarxiv icon

LibAUC: A Deep Learning Library for X-Risk Optimization

Add code
Jun 05, 2023
Figure 1 for LibAUC: A Deep Learning Library for X-Risk Optimization
Figure 2 for LibAUC: A Deep Learning Library for X-Risk Optimization
Figure 3 for LibAUC: A Deep Learning Library for X-Risk Optimization
Figure 4 for LibAUC: A Deep Learning Library for X-Risk Optimization
Viaarxiv icon

Provable Multi-instance Deep AUC Maximization with Stochastic Pooling

Add code
May 18, 2023
Viaarxiv icon

Benchmarking Deep AUROC Optimization: Loss Functions and Algorithmic Choices

Add code
Mar 29, 2022
Figure 1 for Benchmarking Deep AUROC Optimization: Loss Functions and Algorithmic Choices
Figure 2 for Benchmarking Deep AUROC Optimization: Loss Functions and Algorithmic Choices
Figure 3 for Benchmarking Deep AUROC Optimization: Loss Functions and Algorithmic Choices
Figure 4 for Benchmarking Deep AUROC Optimization: Loss Functions and Algorithmic Choices
Viaarxiv icon

When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee

Add code
Mar 04, 2022
Figure 1 for When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee
Figure 2 for When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee
Figure 3 for When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee
Figure 4 for When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee
Viaarxiv icon

A Unified DRO View of Multi-class Loss Functions with top-N Consistency

Add code
Dec 30, 2021
Figure 1 for A Unified DRO View of Multi-class Loss Functions with top-N Consistency
Figure 2 for A Unified DRO View of Multi-class Loss Functions with top-N Consistency
Figure 3 for A Unified DRO View of Multi-class Loss Functions with top-N Consistency
Viaarxiv icon