Shuai Tang

I help machines learn something useful, hopefully. 🤷‍♂️ professional procrastinator & mediocre singer. not straight.

[Shuai Tang @Shuai93Tang]

@_rdgao segmentation fault

[Shuai Tang @Shuai93Tang]

@bradleyvoytek @crocodoyle @neuromatch short meetings are great so far 😆

[Shuai Tang @Shuai93Tang]

@tallinzen I somehow think that nowadays people only need to put their work online, either arxiv or twitter, and that's enough for people to claim the credits of their work. At the end of the day, the originality of the idea matters the most I guess.

[Shuai Tang @Shuai93Tang]

feel like i didn't take my first day of job srsly. manager wore a shirt, and i was in my pjs.

[Shuai Tang @Shuai93Tang]

@CaseyNewton does that mean I should put pfizer there to get more matches

[Shuai Tang @Shuai93Tang]

@ericjang11 confs provide free registrations to top reviewers, which i think serves as some kinda incentive for GRAD STUDENTS to write better reviews. I honestly don't think it matters to anyone else.

[Shuai Tang @Shuai93Tang]

well, my special talent is that I can tune the baseline to be f*cking strong, and then whatever I came up with can't beat it anymore :shrug

[Steve @Zergfriend]

Can someone explain this weird phenomenon I see in research papers?

[Shuai Tang @Shuai93Tang]

@KevinKaichuang @Werdnamai @andrewgwils @pgmoren @adamianou something alone the line of Fisher kernel. https://t.co/UDwNM3RmJZ

[Shuai Tang @Shuai93Tang]

@_rdgao ya silly b 🤪🤪🤪

[Shuai Tang @Shuai93Tang]

Yoo, I defended my thesis and graduated! Well, that happened last month actually.

[Shuai Tang @Shuai93Tang]

coauthors made this pic to illustrate the high-level idea :D @ wesley (well, he is not twitter) @andrewgwils @pgmoren @adamianou

[Shuai Tang @Shuai93Tang]

Check out our new paper on linearising neural networks for fast adaptation. paper: https://t.co/bOZ3TwJeNP code: https://t.co/jarMuQJNDq

[Shuai Tang @Shuai93Tang]

the hardest part is to make the linear system efficient. wesley made it possible through his novel implementation of scalable Fisher vector product

[Shuai Tang @Shuai93Tang]

Okay, I mean, there is also boosting algorithms and random forests

[Shuai Tang @Shuai93Tang]

Although I work on deeplearning rated things, I’m still quite surprised and impressed by how much deeplearning has been applied in various domains and in services we use everyday, and by the effort on making it scalable with tolerable latency

[Shuai Tang @Shuai93Tang]

@EricLeonardis Well, missed a conjunction there 🤦‍♂️

[Shuai Tang @Shuai93Tang]

@EricLeonardis Hegel’s arguments on China not having philosophical history or Chinese philosophy not worth being studied are still very much preserved in philosophy departments, we can’t diminish Hegel’s early contribution on reformalising idealism based on meta-physics beings...

[Shuai Tang @Shuai93Tang]

Drowning in voluntary reviewing duties for a few confs. At least there is a piece of good news this morning.