Zi Wang, Ph.D. @ziwphd Twitter profile

Last Seen Profiles

@gogoo_1406

@eustacewinner

@stwmaniax

@stw_pdg

@stwmaniax

@SBLiveOK

@stwmaniax

@gerthansen08

@dmv_dxxx

@loafboyjayy

@Brooklyn304CL

@bokeplokalmalam

@BinorRaja

@elnema_ahmed

@alirezaopmc

@minamim__m

@Castle_G69

@cyuncyun_cyunko

@yuuhirano0207

@shaunattwood

@tokl_tayl

@kopsirgenreyhan

@vinayrkrishnan

@stwmaniax

@kmh

@Binor_stw69

@Qq6223

@IraqSexKareemah

@chebura_dance

@stw46

@RohanNadkarni

@P0WCHITAS

@eptals

@boneslingernyc

@abdulmajeedsu

Zi Wang, Ph.D.

@ziwphd

2 years

Interested in Bayesian optimization? Check out Google BayesOpt Speaker Series! We’ve had a fascinating series of talks given by speakers including @arkrause , Yoshua Bengio, @peter_i_frazier , Kirthevasan Kandasamy, @PhilippHennig5 .. Stay tuned for more!

Google BayesOpt Speaker Series 2021-2022

www.youtube.com

1

62

342

Zi Wang, Ph.D.

@ziwphd

6 years

Still don’t believe reinforcement learning is an appropriate model for human-level intelligence. We are taught delayed gratitude all our lives while RL agents use discounted reward?! Like wut?! I’ll be so depressed if I’m an RL agent.

11

14

180

Zi Wang, Ph.D.

@ziwphd

2 years

Alex Terenin's new Google TechTalk on Gaussian processes is a must-watch! @avt_im He offers some really cool perspectives on this statistical tool. I learned a lot from it, and I'm sure you will too!

Pathwise Conditioning and Non-Euclidean Gaussian Processes

A Google TechTalk, presented by Alexander Terenin, 2022-11-28BayesOpt Speaker Series. ABSTRACT: In Gaussian processes, conditioning and computation of poste...

www.youtube.com

3

28

162

Zi Wang, Ph.D.

@ziwphd

10 months

Our team at @GoogleDeepMind is hiring a student researcher based in Cambridge MA in 2024 to work on uncertainty-aware decision making for LLMs and/or text-to-image models. Check out the full info and how to apply:

1

17

153

Zi Wang, Ph.D.

@ziwphd

6 years

@Tkaraletsos @roydanroy I had an even funnier review from Nips before: what’s that little square in the end of the proof

2

1

49

Zi Wang, Ph.D.

@ziwphd

5 years

#defended

3

0

45

Zi Wang, Ph.D.

@ziwphd

6 years

Very fancy lunch at MIT ML retreat. Grad students can’t believe they are having cup cakes with the cup made of actual chocolate. Formal sit down lunch rather than sandwich box. Mind blown.

0

2

47

Zi Wang, Ph.D.

@ziwphd

2 years

Jasper @latentjasper talking about the ongoing journey towards BIG Gaussian processes! A team effort with @hoonkp , Ben Adlam, @shreyaspadhy and @zacharynado . Join us at NeurIPS GP workshop

0

6

43

Zi Wang, Ph.D.

@ziwphd

2 years

Super interesting talk by @ilijabogunovic on robust Bayesian optimization! View it at our BayesOpt Speaker Series on @Google TechTalks:

Robust Design Discovery and Exploration in Bayesian Optimization

A Google TechTalk, presented by Ilija Bogunovic, 2022/10/04BayesOpt Speaker Series - ABSTRACT: Whether in biological design, causal discovery, material produ...

www.youtube.com

0

9

34

Zi Wang, Ph.D.

@ziwphd

10 months

Introducing Gaussian Process Probes (GPP) for probing and measuring uncertainty about concepts represented by AI models. #NeurIPS poster #1504 this Tue at 5pm! Paper: Joint work with @alex_y_ku , @jasonbaldridge , Tom Griffiths @cocosci_lab and @_beenkim

3

8

32

Zi Wang, Ph.D.

@ziwphd

2 years

Interpretable BayesOpt solution via sparsity! Nice talk by @su_lin_liu at NeurIPS GP workshop!

0

1

26

Zi Wang, Ph.D.

@ziwphd

2 years

Where should Bayesian priors come from and how can we trust them for decision making? tldr is pre-training to align with expert beliefs! Kudos to @GeorgeEDahl , @kswersk , Chansoo Lee, @zacharynado , @jmgilmer , @latentjasper , @ZoubinGhahrama1 for the collab!

Pre-trained Gaussian Processes for Bayesian Optimization

Bayesian optimization (BO) has become a popular strategy for global optimization of expensive real-world functions. Contrary to a common expectation that BO is suited to optimizing black-box...

arxiv.org

Google AI

@GoogleAI

2 years

Hyper Bayesian optimization (HyperBO) is a highly customizable interface that pre-trains a Gaussian process model and automatically defines model parameters, making Bayesian optimization easier to use while outperforming traditional methods. Learn more →

10

132

643

0

2

14

Zi Wang, Ph.D.

@ziwphd

10 months

Using LLMs for a new domain-specific language? We developed Grammar Prompting to teach them rules&structures, boosting their skills in semantic parsing, planning 🤖, and even molecule generation! Excited to share this work #NeurIPS2023 poster #329 Tuesday at 5pm!

Bailin Wang

@bailin_28

10 months

@Ziwphd and I will present our Grammar Prompting work tomorrow at #NeurIPS2023 Happy to chat more on - LLMs for structured language generation (e.g., programs, molecules and robotic plans) - Structured/formal chain-of-thought for reasoning

1

3

20

0

2

13

Zi Wang, Ph.D.

@ziwphd

2 years

Join us at Room 387 for NeurIPS workshop on Gaussian processes! @ZoubinGhahrama1 giving a opening remark now

0

3

13

Zi Wang, Ph.D.

@ziwphd

2 years

InfoBax! @willieneis on general information theoretic framework for algorithm execution @ NeurIPS GP workshop

0

3

12

Zi Wang, Ph.D.

@ziwphd

8 months

Ever struggled with transferring knowledge in Bayesian optimization due to varying dimensions? Our TMLR paper solves this with MPHD! New Gaussian process Model Pre-training method works seamlessly across Heterogeneous Domains. #MachineLearning

Transfer Learning for Bayesian Optimization on Heterogeneous Search Spaces

Bayesian optimization (BO) is a popular black-box function optimization method, which makes sequential decisions based on a Bayesian model, typically a Gaussian process (GP), of the function. To...

arxiv.org

3

20

Zi Wang, Ph.D.

@ziwphd

2 years

Check out to learn more about how to pre-train a Gaussian process!

GitHub - google-research/hyperbo: Pre-trained Gaussian processes for Bayesian optimization

Pre-trained Gaussian processes for Bayesian optimization - google-research/hyperbo

github.com

Kevin Swersky

@kswersk

2 years

This is a really natural framework to improve Bayesian optimization when you have access to related optimization tasks Joint work with @ziwphd , @GeorgeEDahl , Chansoo Lee, @zacharynado , @jmgilmer , @latentjasper , @ZoubinGhahrama1

1

5

27

0

1

11

Zi Wang, Ph.D.

@ziwphd

2 years

Awesome talk by @anmakaro from ETH Zurich on Bayesian Optimization in the Wild: Risk-Averse Decisions and Budget Constraints, featuring best paper award at #AutoML_Conf . Check it out on Google BayesOpt @ YouTube! #bayesopt #automl

Bayesian Optimization in the Wild: Risk-Averse Decisions and Budget...

A Google TechTalk, presented by Anastasia Makarova, 2022/08/23Google BayesOpt Speaker Series - ABSTRACT: Black-box optimization tasks frequently arise in hig...

www.youtube.com

0

1

9

Zi Wang, Ph.D.

@ziwphd

11 months

Do we really need a gigantic amount of data to train AI models? We've been exploring this question from several facets and here's our first step to ensure robustness during data downsampling!

Ilija Bogunovic

@ilijabogunovic

11 months

How to downsample data to safeguard against worst-case generalization performance issues (stemming from class imbalance and distributional shifts)? Great work with @bankes_william , G. Hughes and @ziwphd

0

1

7

0

1

7

Zi Wang, Ph.D.

@ziwphd

2 years

Never thought I would ever say this, but look.. transformers can do optimization, quite well 😬 glad to be involved in this project.

Google AI

@GoogleAI

2 years

Today on the blog learn about the OptFormer, one of the first Transformer-based frameworks for hyperparameter tuning, learned from large-scale optimization data using flexible text-based representations →

10

115

457

0

6

Zi Wang, Ph.D.

@ziwphd

2 years

Special thanks to our video producer Mark Chow and our organizing/advisory committee @latentjasper , Chansoo Lee, @SagiPerel and @ZoubinGhahrama1 !

0

1

5

Zi Wang, Ph.D.

@ziwphd

8 months

Thank you to those who expressed interest in the position. I've received hundreds of emails and unfortunately won't have the bandwidth to reply. Please make sure to apply through the official channel directly

Careers

At the heart of Google DeepMind’s mission is our commitment to act as responsible pioneers in the field of AI, in service of society’s needs and expectations.

deepmind.google

Zi Wang, Ph.D.

@ziwphd

10 months

Our team at @GoogleDeepMind is hiring a student researcher based in Cambridge MA in 2024 to work on uncertainty-aware decision making for LLMs and/or text-to-image models. Check out the full info and how to apply:

1

17

153

1

0

5

Zi Wang, Ph.D.

@ziwphd

8 months

MPHD enhances Bayesian optimization with domain adaptability. Using a neural net pre-trained in existing domains, MPHD generates tailored Gaussian process models for new domains, leading to smarter optimization decisions.

0

3

Zi Wang, Ph.D.

@ziwphd

2 years

Super cool talk by Frank Hutter at the latest Google BayesOpt Speaker Series: “Deep Learning 2.0: How Bayesian Optimization May Power the Next Generation of DL”. A new era of deep learning? Thank you @FrankRHutter for joining us!

Deep Learning 2.0: How Bayesian Optimization May Power the Next...

A Google TechTalk, presented by Frank Hutter, 2022/6/14ABSTRACT: BayesOpt TechTalk Series. Deep Learning (DL) has been incredibly successful, due to its abil...

www.youtube.com

0

3

Zi Wang, Ph.D.

@ziwphd

8 months

Congrats to my wonderful co-authors Zhou Fan @evensgn & Nicole Han @nahelocin from @Harvard for the paper acceptance!

1

0

3

Zi Wang, Ph.D.

@ziwphd

8 months

@evensgn @nahelocin @Harvard Our paper reveals a fundamental connection: multi-task learning and single-task learning are equivalent for stationary kernels. This theoretical insight means MPHD models are guaranteed to improve over time, ultimately reflecting the real-world process that produces the data.

0

2

4

Zi Wang, Ph.D.

@ziwphd

10 months

@alex_y_ku @jasonbaldridge @cocosci_lab @_beenkim GPP judges uncertainty rationally, just like a rational human, e.g. no extreme judgements if no evidence, and getting close to a firm judgment only if episteme is high. We can even prove theoretically that episteme has no upper bound, i.e., there is no limit to knowledge.

1

0

2

Zi Wang, Ph.D.

@ziwphd

10 months

With episteme, GPP knows when it does not know. This means GPP not only probes, but also detects out-of-distribution data! On 3D Shapes and ImageNet OOD detection tasks, GPP performed consistently well and on par with the best benchmark method for each task.

0

1

Zi Wang, Ph.D.

@ziwphd

10 months

@alex_y_ku @jasonbaldridge @cocosci_lab @_beenkim GPP probes model representation with epistemic and aleatory uncertainty evaluated as episteme (knowledge / confidence), alea (intrinsic randomness) and judged probability. This helps GPP distinguish fuzzy concepts and unknown concepts, an ability linear probes do not possess.

1

0

1

Zi Wang, Ph.D.

@ziwphd

2 years

Contributed Talk by Renzhi Chen on hyperparameter hardware/software codesign! Join us at NeurIPS GP workshop

0

1