paper-with-me

홈 › Papers

Differentiable Weight Masks for Domain Transfer

2023-08-26 · Samar Khanna, Skanda Vaidyanath, Akash Velu

One of the major drawbacks of deep learning models for computer vision has been their inability to retain multiple sources of information in a modular fashion. For instance, given a network that has been trained on a source task, we would like to re-train this network on a similar, yet different, target task while maintaining its performance on the source task. Simultaneously, researchers have extensively studied modularization of network weights to localize and identify the set of weights culpable for eliciting the observed performance on a given task. One set of works studies the modularization induced in the weights of a neural network by learning and analysing weight masks. In this work, we combine these fields to study three such weight masking methods and analyse their ability to mitigate "forgetting'' on the source task while also allowing for efficient finetuning on the target task. We find that different masking techniques have trade-offs in retaining knowledge in the source task without adversely affecting target task performance.

📄 PDF Abstract BibTeX arXiv:2308.13957

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

CWP: Instance complexity weighted channel-wise soft masks for network pruning

2022-09-08 · Jiapeng Wang, Ming Ma, Zhenhua Yu

Existing differentiable channel pruning methods often attach scaling factors or masks behind channels to prune filters with less importance, and implicitly assume uniform contribution of input samples to filter importanc…

Network Pruning

Piggyback: Adapting a Single Network to Multiple Tasks by Learning to Mask Weights

2018-01-19 · ECCV 2018 9 · Arun Mallya, Dillon Davis, Svetlana Lazebnik

This work presents a method for adapting a single, fixed deep neural network to multiple tasks without affecting performance on already learned tasks. By building upon ideas from network quantization and pruning, we lear…

Continual LearningQuantization

MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models

2024-09-26 · Gongfan Fang, Hongxu Yin, Saurav Muralidharan, Greg Heinrich 외

Large Language Models (LLMs) are distinguished by their massive parameter counts, which typically result in significant redundancy. This work introduces MaskLLM, a learnable pruning method that establishes Semi-structure…

Large Language ModelModel CompressionNetwork PruningTransfer Learning

Non-Differentiable Supervised Learning with Evolution Strategies and Hybrid Methods

2019-06-07 · Karel Lenc, Erich Elsen, Tom Schaul, Karen Simonyan

In this work we show that Evolution Strategies (ES) are a viable method for learning non-differentiable parameters of large supervised models. ES are black-box optimization algorithms that estimate distributions of model…

text-to-speechText to Speech

Are Neural Nets Modular? Inspecting Functional Modularity Through Differentiable Weight Masks

2020-10-05 · ICLR 2021 1 · Róbert Csordás, Sjoerd van Steenkiste, Jürgen Schmidhuber

Neural networks (NNs) whose subnetworks implement reusable functions are expected to offer numerous advantages, including compositionality through efficient recombination of functional building blocks, interpretability, …

Systematic Generalization