上式是总的学习目标,表示了在源域的第 m 层和目标域第 n 层之间迁移多少。分别对应 what 和 where 两个问题。 What to transfer 对于给定的S^m(x)与T_\theta^n(x),我们给予不同的feature map不同的可训练权重,越重要的 feature map 所对应的权重越大,这使得对应的loss也就越受到关注: 图源三元师兄 w...
Learning What and Where to Transfer 概述 这是一篇来自 ICML 2019 的迁移学习论文。作者针对异构师生网络的知识迁移任务,提出了一种基于元学习的迁移学习方法,自动地学习源网络中什么知识需要迁移、迁移到目标网络的什么地方。也就是说,通过元学习来决定: (a)源
本文主要总结了ICML2019的关于meta-learning的文章。 Learning What and Where to Transfer motivation:迁移学习一个常用的范式是在source dataset上训练一个模型,在target dataset上进行fine-tine。但是这样无法适用于source和target的任务差异较大,并且source和target的网络结构不一致的情形。本文提出了利用meta-learning的...
Train L2T-ww You can train L2T-ww models with the same settings in our paper. python train_l2t_ww.py --dataset cub200 --datasplit cub200 --dataroot /data/CUB_200_2011 python train_l2t_ww.py --dataset dog --datasplit dog --dataroot /data/dog python train_l2t_ww.py --dataset...
(CoRL2020)DIRL: Domain-Invariant Representation Learning Approach for Sim-to-Real Transfer 论文笔记 本文针对的问题是无监督领域自适应和半监督领域自适应问题。 与传统的对抗领域自适应方法对比,其创新性在于 在对齐边缘概率分布的同时也对齐条件概率分布(虽然感觉现在大家都在对齐条件概率分布,应该不能算新颖了) ...
Learning What and Where to Transfer(ICML 2019)中提出基于meta-learning的迁移学习方法,利用meta-learning学习什么样的迁移策略能够达到最优效果。首先,本文的迁移方法采用了FITNETS: HINTS FOR THIN DEEP NETS(ICLR 2015)提出的思路,在finetune阶段通过对target模型参数和pretrain模型参数添加L2正则化损失,来控制target...
7Learning classifiers for target domain with limited or no labels (paper)ICML 2019codezero/few shot learning 6Learning What and Where to Transfer (paper)ICML 2019meta-TLnew trend 5On Learning Invariant Representation for Domain Adaptation (paper)ICML 2019theory ...
Episodic Memory in Lifelong Language Learning 2019 NeurIPS Continual Learning with Tiny Episodic Memories 2019 ICML Efficient lifelong learning with A-GEM 2019 ICLR Learning to Learn without Forgetting by Maximizing Transfer and Minimizing Interference 2019 ICLR Large Scale Incremental Learning 2019 CVPR On...
Learning What and Where to Transfer(ICML 2019)中提出基于meta-learning的迁移学习方法,利用meta-learning学习什么样的迁移策略能够达到最优效果。首先,本文的迁移方法采用了FITNETS: HINTS FOR THIN DEEP NETS(ICLR 2015)提出的思路,在finetune阶段通过对target模型参数和pretrain模型参数添加L2正则化损失,来控制target...
De novo protein design—From new structures to programmable functions Tanja Kortemme Cell 187.3 (2024) Generative models for protein structures and sequences Chloe Hsu, Clara Fannjiang & Jennifer Listgarten Nat Biotechnol 42, 196–199 (2024) What does it take for an ‘AlphaFold Moment’ in func...