LGCVApr 9, 2024

GO4Align: Group Optimization for Multi-Task Alignment

Tsinghua
arXiv:2404.06486v216 citationsh-index: 16NIPS
Originality Incremental advance
AI Analysis

This addresses task imbalance in multi-task learning, which is an incremental improvement over existing methods.

The paper tackles task imbalance in multi-task learning by proposing GO4Align, an approach that aligns optimization across tasks using adaptive group risk minimization, achieving performance superiority with lower computational costs on diverse benchmarks.

This paper proposes \textit{GO4Align}, a multi-task optimization approach that tackles task imbalance by explicitly aligning the optimization across tasks. To achieve this, we design an adaptive group risk minimization strategy, comprising two techniques in implementation: (i) dynamical group assignment, which clusters similar tasks based on task interactions; (ii) risk-guided group indicators, which exploit consistent task correlations with risk information from previous iterations. Comprehensive experimental results on diverse benchmarks demonstrate our method's performance superiority with even lower computational costs.

Code Implementations1 repo
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes