activity
20202026
most citedTRGP: Trust Region Gradient Projection for Continual Learning

22 citations · 36 across the 5 of their papers we have counts for

collaborators
Showing cs.LGShow all

7 papers · 1 filter

cs.LG2026

Aurora: A Leverage-Aware Spectral Optimizer

Alec Dewulf, Dhruv Pai, Li Yang +2

We show that for tall matrix parameters, like projection matrices in the MLP layers, the Muon update can have row norms that are arbitrarily non-uniform. This can lead to a self-re…

cs.LG2026

Dominant-Layer ZO: A Single Layer Dominates Zeroth-Order Fine-Tuning of LLMs

Wanhao Yu, Ziyan Wang, Zheng Wang +7

Zeroth-order (ZO) optimization enables memory-efficient fine-tuning of large language models (LLMs) using only forward passes, but it remains unclear how useful adaptation is distr…

cs.LG20227 cited

Beyond Not-Forgetting: Continual Learning with Backward Knowledge Transfer

Sen Lin, Li Yang, Deliang Fan +1

By learning a sequence of tasks continually, an agent in continual learning (CL) can improve the learning performance of both a new task and `old' tasks by leveraging the forward k…

cs.LG202222 cited

TRGP: Trust Region Gradient Projection for Continual Learning

Sen Lin, Li Yang, Deliang Fan +1

Catastrophic forgetting is one of the major challenges in continual learning. To address this issue, some existing methods put restrictive constraints on the optimization space of…

cs.LG20213 cited

GROWN: GRow Only When Necessary for Continual Learning

Li Yang, Sen Lin, Junshan Zhang +1

Catastrophic forgetting is a notorious issue in deep learning, referring to the fact that Deep Neural Networks (DNN) could forget the knowledge about earlier tasks when learning ne…

cs.LG20213 cited

RA-BNN: Constructing Robust & Accurate Binary Neural Network to Simultaneously Defend Adversarial Bit-Flip Attack and Improve Accuracy

Adnan Siraj Rakin, Li Yang, Jingtao Li +5

Recently developed adversarial weight attack, a.k.a. bit-flip attack (BFA), has shown enormous success in compromising Deep Neural Network (DNN) performance with an extremely small…