@WRuohan

ML Research Scientist, Institute for Infocomm Research, A*STAR, ex-@AWS

 Singapore
Joined June 2016
  • Tweets14
  • Following160
  • Followers48
terrytao.wordpress.com/2026/…

A Severe Misalignment of AI in Mathematics

I am proud to be among the list of 25 initial signatories — all Fields Medallists — to the declaration below, which grew out of discussions between ourselves over the last week. We have…

terrytao.wordpress.com
27
Happy that our paper "Robust Meta-Representation Learning via Global Label Inference and Classification" with Isak Falk, Massi Pontil and Carlo Ciliberto has been accepted to PAMI. check it out at computer.org/csdl/journal/tp… (1/3)

Robust Meta-Representation Learning via Global Label Inference and Classification

Few-shot learning (FSL) is a central problem in meta-learning, where learners must efficiently learn from few labeled examples. Within FSL, feature pre-training has become a popular strategy to...

computer.org
1
362
The paper builds upon our previous work "The Role of Global Labels in Few-Shot Classification and How to Infer Them" from Nuerips 2021. We show that standard pre-training is a special instance of meta-learning, but with improved convergence rate. (2/3)
1
62
The results in turn provide a theoretical perspective towards why pre-trained models are robust few-shot learners. #fewshotlearning #pretraining (3/3)
14
Our paper "The Role of Global Labels in Few-Shot Classification and How to Infer Them" with Massi Pontil and Carlo Ciliberto has been accepted to #NeurIPS2021. Check it out at arxiv.org/abs/2108.04055

The Role of Global Labels in Few-Shot Classification and How to Infer Them

Few-shot learning is a central problem in meta-learning, where learners must quickly adapt to new tasks given limited training data. Recently, feature pre-training has become a ubiquitous...

arxiv.org
1
1
1
Our paper "Structured Prediction for Conditional #MetaLearning" with @YiannisDemiris and Carlo Ciliberto has been accepted to #NeurIPS2020. We will release the updated version of the paper soon. An earlier version is at arxiv.org/abs/2002.08799.

Structured Prediction for Conditional Meta-Learning

The goal of optimization-based meta-learning is to find a single initialization shared across a distribution of tasks to speed up the process of learning new tasks. Conditional meta-learning seeks...

arxiv.org
7
Our paper Random Expert Distillation: Imitation Learning via Expert Policy Support Estimation (arxiv.org/abs/1905.06750) has been accepted to ICML 2019. We propose a new take on imitation learning without adversarial training of the reward function.
1
2
9
Our paper Real-Time Workload Classification during Driving using HyperNetworks (arxiv.org/abs/1810.03145) has been accepted for IROS 2018. Our proposed architecture is designed to tackle noisy and dynamic data like gaze patterns.

Real-Time Workload Classification during Driving using HyperNetworks

Classifying human cognitive states from behavioral and physiological signals is a challenging problem with important applications in robotics. The problem is challenging due to the data...

arxiv.org
1
3
MAGAN: Margin Adaptation for Adversarial Networks, our new paper on auto-encoder based GAN. arxiv.org/abs/1704.03817 github.com/RuohanW/magan

GitHub - RuohanW/magan: Tensorflow Implementation for MAGAN: Margin Adaptation for Adversarial...

Tensorflow Implementation for MAGAN: Margin Adaptation for Adversarial Networks - RuohanW/magan

github.com
1
1
3
Why large-batch learning fails. arxiv.org/abs/1609.04836
Load more