BriefGPT.xyz
Apr, 2023
借助剪枝辅助的特定领域权重调节实现持续域适应
Continual Domain Adaptation through Pruning-aided Domain-specific Weight Modulation
HTML
PDF
Prasanna B, Sunandini Sanyal, R. Venkatesh Babu
TL;DR
本文提出了一种方法,旨在解决连续学习中的无监督域适应问题,通过修剪实现框架来保留特定于域的知识,并使用一种基于批次标准化的度量方法进行有效推理,取得了良好的性能,同时在防止过去领域的灾难性遗忘方面显著改善。
Abstract
In this paper, we propose to develop a method to address
unsupervised domain adaptation
(UDA) in a practical setting of
continual learning
(CL). The goal is to update the model on continually changing domains whi
→