Zhao Song

[Ruizhe Zhang @zrz96zrz]

⁦@realZhaoSong⁩ https://t.co/nekijs5tvS

[Zhao Song @realZhaoSong]

https://t.co/6BDR9nO1MK

[Beidi Chen @BeidiChen]

Can sparse training achieve wall-clock time speed up on GPU?
Yes! Simple and static #sparsity -> 2.5x faster🚀 training MLP-Mixer, ViT, and GPT-2 medium from scratch with NO drop in accuracy.
https://t.co/zpbLkiWk0o (#NeurIPS2021)
https://t.co/QbPtFEiGSy [1/6]

[Sanjeev Arora @prfsanjeevarora]

InstaHide: trains deep nets on encrypted data only. Very fast, preserves privacy of user data, small accuracy loss (unlike differential privacy).

[Zhao Song @realZhaoSong]

Very beautiful work on the parallel shortest path!

[Zhao Song @realZhaoSong]

I just went back to Princeton for a class and now I am back to Simons for the data science reunion! What a hectic week!

[Zhao Song @realZhaoSong]

I finally go back to the Simons Institute for the reunion of the Lower Bounds Program! Cannot wait to learn more about GCT!

Also, stay tuned for my talk tomorrow!

[Zhao Song @realZhaoSong]

I am sure this will be fantastic

[Zhao Song @realZhaoSong]

I finally become a doctor! Now it is the time to tweet more.

[Zhao Song @realZhaoSong]

Thanks so much, Gautam!

[Gautam Kamath @thegautamkamath]

Congratulations to the inimitable @realZhaoSong on successfully defending his PhD thesis!

[Zhao Song @realZhaoSong]

Happy to see that our paper is mentioned in a GCT talk!

[Zhao Song @realZhaoSong]

Learned a lot about GCT from the Algebraic Methods workshop in Simons. GCT is so amazing!

[Zhao Song @realZhaoSong]

Had a WONDERFUL night!

[Zhao Song @realZhaoSong]

My computer stops working for no reasons. It seems like I am destined to attend the party tonight.

[Zhao Song @realZhaoSong]

Just gave a (very serious) talk at the Sublinear Algorithms and Nearest-Neighbor Search workshop at Simons.

[Zhao Song @realZhaoSong]

@hoonoseme Yeah, you are right, I have to make all my coauthors happy! Sorry I have to miss the party.

[Zhao Song @realZhaoSong]

Heard that my housemates are going to hold an exponential size theory party at home, hope I can make it.

[Gautam Kamath @thegautamkamath]

An arms race between Zeyuan and @ilyaraz2 at @MSFTResearch

[Zhao Song @realZhaoSong]

Heard that the air quality at Berkeley is close to BEIJING! Glad that I went back to Austin for a short break recently.

[Sanjeev Arora @prfsanjeevarora]

Hoping to read new papers by Allen-Zhu et al. Training provably converges on greatly overparametrized deep nets. And such overparametrized deep nets can generalize when trained on data from teacher net. https://t.co/hv5DAxsApp and https://t.co/kmns5fQ43w