vikas verma

Deep Learning

[Yann N. Dauphin @ynd]

📣 Excited to announce the First Workshop on Interpolation Regularizers such as Mixup at @NeurIPSConf.

Paper submission deadline: September 22, 2022

Speakers: @chelseabfinn, @prfsanjeevarora, Kenji Kawaguchi, Youssef Mroueh, Alex Lamb.

https://t.co/wf48y2uECM

🧵👇

[hardmaru @hardmaru]

Nvidia Reveals Its CEO Was Computer Generated in Keynote Speech
https://t.co/kWE8h5APpz

[Moshe Vardi @vardi]

Nerd joke :-)

[vikas verma @vikasverma1077]

@lmthang @quocleix @hieupham789 thanks for sharing @lmthang!

[Thang Luong @lmthang]

Glad to share our #ICML2021 work, DACL, a domain-agnostic contrastive learning approach that works across domains (tabular, images, graphs)! DACL uses Mixup to create pos/neg examples with theoretical foundation.
Paper: https://t.co/Xivu0VkvRw
ICML Poster: https://t.co/BHeG4XVBpu

[vikas verma @vikasverma1077]

If you work on ML applications that require better node classification accuracy without additional memory or computation overhead, you might like this paper!

[vikas verma @vikasverma1077]

Check out our paper at AAAI2021 https://t.co/9X0JUiGhhs

Investigates how to improve the performance of existing GNNs using hidden state mixing, with theoretical analysis of the proposed method.

[Simon Guiroy @GuiroySimon]

Our paper is at #ICML2020 - Lifelong Learning Workshop!

Do we need labeled data from held-out classes for model selection in Few-shot Meta-Learning? Our method doesn’t. Using all available data for training, results show it can outperform meta-validation.
https://t.co/LBJrG9LmVr

[hardmaru @hardmaru]

Vim finally usable on a smartphone.

[vikas verma @vikasverma1077]

@chris_j_beckham and Sina Honari presenting our work on Adversarial Mixup Resynthesizer at @NeurIPSConf @devon_hjelm, Farnoosh Ghadiri, Alex Lamb @chrisjpal

[Christopher Beckham, PhD @chris_j_beckham]

@devon_hjelm @vikasverma1077 @chrisjpal New version finally out on arxiv (post-submission but pre-camera-ready), a whole lot has changed since then! https://t.co/LSUvimNUBB

[vikas verma @vikasverma1077]

Thanks @ykilcher for the super nice video presentation!

[Yannic Kilcher 🇸🇨 @ykilcher]

Manifold Mixup: Better Representations by Interpolating Hidden States
https://t.co/5UCYlkpSeo
@vikasverma1077

[vikas verma @vikasverma1077]

@dendisuhubdy 😅

[vikas verma @vikasverma1077]

Me vs reviewers these days !

[Simon Guiroy @GuiroySimon]

Also at "ICML Workshop 2019: Understanding and Improving Generalization in Deep Learning” #icml2019 . Together with @vikiverma and @chrisjpal

[vikas verma @vikasverma1077]

@dendisuhubdy I am the orange one :P

[vikas verma @vikasverma1077]

Also check the follow-up paper by Google Research with even better results: https://t.co/S7sQK0bw4R

[vikas verma @vikasverma1077]

Check out our IJCAI 2019 paper on SSL: https://t.co/lXp0UnpGjE. Turns out adversarial perturbations are not necessary: perturbations in the direction of other samples does the job even better.Paper extends Mixup to SSL (w\ Alex, Juho Kannala, David Lopez-Paz and Yoshua Bengio)

[Thomas G. Dietterich @tdietterich]

Dear academic colleagues. It is not appropriate for class projects to be submitted to @arxiv_org unless they are also being submitted for publication. I and my fellow moderators are seeing more and more of this behavior. It wastes our time and the time of your colleagues