Weitong Zhang
@WeitongZhang
Followers
1,307
Following
3,132
Media
22
Statuses
77
Assistant Professor @UNC | CS Ph.D. @UCLA | Ex-intern @nvidia ｜ @amazon fellow
Chapel Hill
Joined April 2020
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
GCF HEADLINER LISA
• 404113 Tweets
Georgia
• 214638 Tweets
#precure
• 74999 Tweets
Primanti Bros
• 71392 Tweets
Parque Lezama
• 68559 Tweets
Martinez
• 66689 Tweets
Román
• 66543 Tweets
PRETTY GIRL MANTRA
• 52570 Tweets
Bama
• 48284 Tweets
#仮面ライダーガヴ
• 38795 Tweets
Travis Hunter
• 37781 Tweets
#ブンブンジャー
• 34520 Tweets
Tuscaloosa
• 30495 Tweets
Heisman
• 28362 Tweets
Auburn
• 25592 Tweets
Riquelme
• 23999 Tweets
Kirby
• 23266 Tweets
#MostRequestedLive
• 21548 Tweets
Milroe
• 18992 Tweets
Saban
• 17185 Tweets
Dawgs
• 14927 Tweets
Jeremiah Smith
• 14827 Tweets
Carson Beck
• 14333 Tweets
#RollTide
• 13829 Tweets
Belgrano
• 13303 Tweets
ショウマ
• 11941 Tweets
Ocee
• 10374 Tweets
Pinned Tweet
Thrilled to announce I've defended my PhD dissertation 🎓today at
@CS_UCLA
! Immense thanks to my advisor
@QuanquanGu
and committee members for their invaluable guidance and insights. It's been a remarkable five-year journey, filled with unforgettable experiences.
📢Starting July
12
2
124
I will be at
#NeurIPS2023
next week to present two workshop papers on
#LLM
for molecule and other topics. Happy chatting with you about 🤖
#RL
, 🔬
#AIforScience
and many!
🎓I’m on the academia job market for 2024. Looking forward to hearing your advice on job search!
See you all
2
5
48
Can Thompson Sampling (TS) strategy probably explore the neural network contextual bandits? Check out our paper in
#ICLR2021
to be presented on May 6th, 5-7 pm PDT.
*Paper page:
*Gathertown: Poster Section 12, Spot C4
1
1
21
Reflecting on this year's
#Thanksgiving
, I'm filled with gratitude. A heartfelt thank you to Prof.
@QuanquanGu
, Prof.
@yayitsamyzhang
, Dr.
@32Eaton
, and my esteemed collaborators, including
@Yihe__Deng
,
@JunkaiZZ
, and others for their invaluable help in my research and job search
1
0
22
Will be presenting our poster MoleculeGPT: Instruction Following Large Language Models for Molecular Property Prediction at the AI4D3 workshop at Neurips tomorrow. See you all in Room 242. The poster session is ⌚️1:20 p.m. - 2:25 p.m.
Paper link 📰
I will be at
#NeurIPS2023
next week to present two workshop papers on
#LLM
for molecule and other topics. Happy chatting with you about 🤖
#RL
, 🔬
#AIforScience
and many!
🎓I’m on the academia job market for 2024. Looking forward to hearing your advice on job search!
See you all
2
5
48
0
3
16
Can Reward-Free Exploration help model-based Reinforcement Learning with linear function approximation? Check out our paper in
#NeurIPS2021
to be presented on Dec. 8th, 4:30 PM — 6:00 PM PST.
* Paper Page
* Gathertown: Poster Session 4, Spot G1
1
1
8
I'm entering the academic job market in 2024. Any leads or advice are greatly appreciated. Please feel free to connect and share!
0
0
5
In this paper, we propose a Neural Thompson Sampling algorithm which attains an O(\sqrt {T}) regret and robust performance in practice. Joint work w/
@DongruoZ
@LihongLi20
@QuanquanGu
0
0
2
@DWUhlfelderLaw
Take care and stay safe! According to our prediction ( and checkout each state), new cases in Florida will still increase in the further several days. So keep social distancing and remember to wear a mask if you go out!
0
0
1
Come and join us in poster session 2
#icml
#ICML2023
in discussing the exploration strategies in RL with long horizons, especially when we are getting rid of the reward functions.
0
0
2
RL with linear function approximation is a well-studied topic, but achieving the optimal regret guarantee in linear MDPs remains an open problem. Excitingly, our work proposes an innovative algorithm (LSVI-UCB++) that provides a solution to this challenge! 🚀 Check it out! (1/2)
1
2
7
0
0
2
TL;DR: We present an interplay between misspecification with the sub-optimality gap in linear bandits, which demonstrates the circumstances in which misspecified linear bandits can be efficiently learned and when they cannot. (2/3)
1
0
2
@ToddGloria
Exactly, according to our prediction ( , click California and select San Diego), the confirmed cases in SD is still increasing. That means there are still many unreported cases there. More tests needed to discover these unreported cases!
0
1
2
Paper link:
Poster session link:
Time: Tue 25 Jul, 11 a.m. HST — 1:30 p.m. HST
Location: Poster Session 1 @ Exhibit Hall 1
#627
A joint work with
@JiafanHe
, Zhiyuan and
@QuanquanGu
(3/3)
0
0
1
@thehill
According to the projection by UCLA statistical machine learning lab
@QuanquanGu
(), the United States will have 130K deaths in early August. Based on these models’ forecasts for US fatality at
@CDC
website, I prefer this result as a more accurate one.
1
0
1
In this work, we provide a reward-free exploration for learning linear mixture MDPs with an O(H⁵d²ε⁻²) sample complexity and could be further improved to O(H⁴d(H + d)ε⁻²). We also provide a lower bound with sample complexity Ω(H²dε⁻²). Joint work with
@DongruoZ
@QuanquanGu
1
0
1
@ShrikeTron
@thehill
@QuanquanGu
@cdc
The pandemic has evolved rapidly that all previous prediction models were skewed by various factors, e.g., early reopening, covid parties, no face masking, etc. BTW, YYG predicted 188,698 deaths for August 4, more than 31K off from the true number 157,482
This is the best performing model, by Youyang Gu
@youyanggu
- the numbers for August 4 are not very uplifting, 188,698 projected deaths in the United States.
0
0
2
1
0
1
@apoorva_nyc
@florian_krammer
@LabTaia
@trvrb
@ualbany
That's true, the spread speed of C19 of NY is low, according to our model (check ). But REMEMBER that over 30k people died in NY state because of C19. So keep social distancing because any mutation of the virus might ruin this immunization.
0
0
1
@SovernNation
Right, more testing should be done prior to reopen the state, according to our prediction (see ), there would be a large uncertainty for the prediction of the new cases next month. This could only be prevented by doing more tests.
1
0
0