If you are going to attend ICLR BAICS workshop and you are interested in the interaction between neuroscience and AI, feel free to chat with me at 2-3pm GMT April 26. Here is my virtual poster session link: baicsworkshop.github.io/program/baics_… #ICLR2020 #BAICS2020

Can backpropagation tell us anything about learning in the brain? In our Nature Rev Neurosci paper, we argue that while the brain probably does not implement BP in its literal form, BP has features that are relevant for understanding biological learning. go.nature.com/2z9CSdM

ICLR 2020 starts tomorrow, to prepare for it, I listed for you all the, accepted continual learning papers! (0/12) #ICLR2020 #ContinualLearning

Love the virtual style and have learned a lot this year. Have seen lots of exciting work this year! I actually find the virtual style make stuff more productive!

We have significantly updated our survey on GANs: arxiv.org/pdf/1906.01529…. Code is also available: github.com/sheqi/GAN_Revi…. Nice piece of work with Qi (Roger) She and tomasward.

I should include this nice summary figure in my last tweet!

I’d like to share my new project: Objax, a new high-level JAX API with a PyTorch-like interface! Objax pursues the quest for the simplest design and code that’s as easy as possible to extend without sacrificing performance. github.com/google/objax objax.readthedocs.io/en/latest/inde… 1/7

good quick tutorial on optimizing your PyTorch code ⏲️: youtube.com/watch?v=9mS1fI… quick summary:

Deep learning is just machine learning lol

What the heck is your broadband service? It is broken at times everyday. Can your engineers make it even worse??? Your internet makes my WFH experience awesome. Thank you !!!!!!!! Virgin Media Ireland Virgin Media ❤️

Nice. Broadband down again. I am going to switch provider.

Video Transformer Network pdf: arxiv.org/pdf/2102.00719… abs: arxiv.org/abs/2102.00719

Do We Really Need Explicit Position Encodings for Vision Transformers? pdf: arxiv.org/pdf/2102.10882… abs: arxiv.org/abs/2102.10882

Exciting for moving step from #CVPR2021 in submission -> To appear in #CVPR2021. Thanks for all ACs, reviewers, and PCs for their hard work! Nice start of the week. Move to the next step! Our preprint and code will be out shortly.

Maybe also plus corrected versions of ICLR and AAAI. Fibonacci confirmed!

Facebook AI has built TimeSformer, a new architecture for video understanding. It’s the first based exclusively on the self-attention mechanism used in Transformers. It outperforms the state of the art while being more efficient than 3D ConvNets for video.ai.facebook.com/blog/timesform…

Our CVPR'21 accepted version "ACTION-Net: Multipath Excitation for Action Recognition" has been out arxiv.org/abs/2103.07372 with provided codes: github.com/V-Sense/ACTION…. Here is the highlight: v-sense.scss.tcd.ie/research/actio…. Co-work with Qi (Roger) She and Aljosa Smolic V-SENSE

Go United Go!!!