Yen-Huan Li @yenhuan_li Twitter profile

Last Seen Profiles

@Nathanrutter_10

@bokeplokalmalam

@wavverse

@hard_athe

@thebrutalstorms

@blackman_rile

@CalicoCutGiving

@Makkop9

@OpmInsos

@DelbridgeJaidyn

@muslimmumsvoice

@vivanco_emilio

@oregon__updates

@DULadyPanthers

@longino_is90618

@mariabreslin

@jerkimof

@aya1020_an

@kevlondon4

@nir_tshuva

@Listen60349770

@bts66368581

@Blevins77Owen

@steveconsultsp1

@Drae_slim

@BigleafNetworks

@SEO_KO_BO_

@BrazilianFren

@twitnitnit

@BILLF0UR1

@WalterAYIN

@hxrryxsatellite

@Beauty1329696

@mrbeatdakitty

@HabibYacin

@JMTamburrino

Yen-Huan Li

@yenhuan_li

2 years

Book draft on sampling and optimization by Yin-Tat Lee and Santosh Vempala.

3

54

371

Yen-Huan Li

@yenhuan_li

3 years

Lecture Notes for Mathematics of Machine Learning by Afonso S. Bandeira & Nikita Zhivotovskiy

2

72

328

Yen-Huan Li

@yenhuan_li

4 years

Just noticed Reviewer 2 said our *theory paper* is irreproducible. Reviewer 2 never disappoints me.

12

17

277

Yen-Huan Li

@yenhuan_li

4 years

Francis Bach's lecture notes on "learning theory from first principles." Interesting selection of topics.

2

40

248

Yen-Huan Li

@yenhuan_li

2 years

I did not expect that ICLR would assign me 5 theory papers to review in 12 days. I should have rejected the reviewer invitation.

7

4

192

Yen-Huan Li

@yenhuan_li

4 years

I thank Dr. Tedros for advertising Taiwan. #ThisAttackComeFromTaiwan

2

15

174

Yen-Huan Li

@yenhuan_li

3 years

We will host Machine Learning Summer School 2021, Taipei during the first three weeks of August. We have an outstanding list of lecturers, talking on a variety of topics in modern machine learning. Free registration for students not presenting posters!

3

47

157

Yen-Huan Li

@yenhuan_li

5 years

Book draft with proofs on RL

1

30

157

Yen-Huan Li

@yenhuan_li

3 years

I will refuse to review for any conference that asks me to finish reviewing seven papers in three weeks...

6

4

129

Yen-Huan Li

@yenhuan_li

4 months

A student told me that he believes there might be a connection between the Fenchel conjugate and the Fourier transform. I replied that if such a connection exists, it might be related to the max-plus algebra. It turns out this is indeed the case! (1/2)

3

8

129

Yen-Huan Li

@yenhuan_li

1 month

==== My recommendations today ==== Bandit convex optimization Fast Convergence of Frank-Wolfe algorithms on polytopes Calibrated Language Models Must Hallucinate (1/n)

4

13

99

Yen-Huan Li

@yenhuan_li

3 years

Both authors and reviewers should focus more on whether a paper provides interesting ideas rather than whether it gives SOTA results.

2

6

93

Yen-Huan Li

@yenhuan_li

3 months

Information theorists' version of online learning.

On Jacob Ziv's Individual-Sequence Approach to Information Theory

This article stands as a tribute to the enduring legacy of Jacob Ziv and his landmark contributions to information theory. Specifically, it delves into the groundbreaking individual-sequence...

arxiv.org

2

12

80

Yen-Huan Li

@yenhuan_li

2 years

More and more papers I reviewed reproduce existing proofs without proper citations, hiding how incremental they are or “for completeness” when asked about the technical novelty.

3

2

79

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== Efficient Sampling on Riemannian Manifolds via Langevin MCMC Interior point methods are not worse than Simplex

1

9

76

Yen-Huan Li

@yenhuan_li

1 year

Theorem. Every paper contains an infinite number of typos. Proof. Every time the authors check the paper, they find at least one more typo.

7

6

73

Yen-Huan Li

@yenhuan_li

2 years

I feel several reviewers are not merely rejecting papers but, more precisely speaking, rejecting research directions. To identify them, just check whether their comments apply to all papers in a research direction or not.

4

1

73

Yen-Huan Li

@yenhuan_li

2 months

==== My recommendations today ==== Isoperimetric inequalities in high-dimensional convex sets One World Optimization Seminar Exact Lipschitz Regularization of Convex Optimization Problems

3

11

72

Yen-Huan Li

@yenhuan_li

1 year

Very complete literature review on gradient descent.

0

12

67

Yen-Huan Li

@yenhuan_li

5 months

==== My recommendations today ==== Handbook of Convergence Theorems for (Stochastic) Gradient Methods Sampling via Gradient Flows in the Space of Probability Measures

0

16

66

Yen-Huan Li

@yenhuan_li

5 months

I did not submit any paper to ICML. But when I read the comments of other reviewers for the papers assigned to me, I feel sympathetic to the authors. The review quality is...

1

0

63

Yen-Huan Li

@yenhuan_li

2 months

==== My recommendations today ==== The Bayesian Learning Rule Quantum chi-squared tomography and mutual information testing (1/n)

2

10

61

Yen-Huan Li

@yenhuan_li

4 years

Two lectures on learning theory by Cesa-Bianchi @ MLSS 2020. 1) 2)

Learning Theory, Part 1

Table of Contents (powered by https://videoken.com)0:00:00 Some Elements of Learning Theory0:03:21 Contents0:12:44 Statistical learning0:23:03 Ingredients0:2...

www.youtube.com

0

9

61

Yen-Huan Li

@yenhuan_li

3 years

Machine Learning Summer School 2021 @Taipei has started. We have about 1700 participants from more than 60 countries!

1

5

57

Yen-Huan Li

@yenhuan_li

4 months

==== My recommendations today ==== An optimal lower bound for smooth convex functions Inexact subgradient methods for semialgebraic functions Optimal tradeoffs for estimating Pauli observables (1/2)

1

2

55

Yen-Huan Li

@yenhuan_li

4 months

==== My recommendations today ==== Some highlights from the history of probability Scalable Projection-Free Optimization Methods via MultiRadial Duality Theory (1/2)

1

7

56

Yen-Huan Li

@yenhuan_li

5 months

==== My recommendations today ==== Generalization Bounds: Perspectives from Information Theory and PAC-Bayes What Should we Trust in Trustworthy Machine Learning (1/2)

1

6

54

Yen-Huan Li

@yenhuan_li

5 months

Nesterov: It’s very easy to get non- convex problems but for me, this means that we didn’t think enough. The final step, to say this problem is solved—this is where we get convex formulations. Non- convexity is just the first step.

“This is an unprecedented overflow”: Why the progress of his field alarms Yurii Nesterov

Mathematician Yurii Nesterov is one of the main actors in the field of Convex Optimization, which is crucial for many applications in Automation. Thanks in part to his breakthroughs, that field is...

nccr-automation.ch

0

5

52

Yen-Huan Li

@yenhuan_li

4 months

==== My recommendations today ==== Faster Convergence of Stochastic Accelerated Gradient Descent under Interpolation Sum-of-Squares Relaxations for Information Theory and Variational Inference (1/n)

2

9

47

Yen-Huan Li

@yenhuan_li

11 months

History about the PL or LPL condition.

0

9

46

Yen-Huan Li

@yenhuan_li

4 years

If we ask the conference ACs to put their names on the published papers they handled, just like journal editors, will the AC quality be improved?

4

0

48

Yen-Huan Li

@yenhuan_li

5 months

==== My recommendations today ==== Majority-of-Three: The Simplest Optimal Learner? Directional Smoothness and Gradient Methods: Convergence and Adaptivity (1/n for a large n...)

1

5

46

Yen-Huan Li

@yenhuan_li

6 months

I have been sharing interesting papers with my students daily for years. Now, I feel it might be beneficial to share my recommendations with more people. The papers chosen reflect my personal research directions and interests, and they are definitely not exhaustive.

1

4

43

Yen-Huan Li

@yenhuan_li

2 years

Proud advisor moment at ALT 2023. The speaker and first author, @chungentsai , is an undergraduate! Our work:

1

44

Yen-Huan Li

@yenhuan_li

1 year

A surprising gem written in 1994! Nemirovski: INFORMATION-BASED COMPLEXITY OF CONVEX PROGRAMMING

1

7

39

Yen-Huan Li

@yenhuan_li

2 years

Bo'az Klartag's course on convexity in high dimensions.

Bo'az Klartag - Convexity in High Dimensions I

October 28, 2022This is the first talk in the Minerva Mini-course of Bo'az Klartag, Weizmann Institute of Science and Princeton's Fall 2022 Minerva Distingui...

www.youtube.com

0

6

35

Yen-Huan Li

@yenhuan_li

1 year

Didn't notice that Chapter 17 of this book provides a great introduction to gradient flow.

Variational Analysis in Sobolev and BV Spaces | SIAM Publications Library

epubs.siam.org

0

6

33

Yen-Huan Li

@yenhuan_li

5 months

==== My recommendations today ==== Optimal First-Order Algorithms as a Function of Inequalities Gaussian Cooling and Dikin Walks: The Interior-Point Method for Logconcave Sampling (1/2)

1

2

34

Yen-Huan Li

@yenhuan_li

1 year

Interesting paper originating from a simple observation.

High-Probability Risk Bounds via Sequential Predictors

Online learning methods yield sequential regret bounds under minimal assumptions and provide in-expectation risk bounds for statistical learning. However, despite the apparent advantage of online...

arxiv.org

0

7

33

Yen-Huan Li

@yenhuan_li

2 years

Spent one whole day on discussions as a reviewer. Haven't seen any reviewer response as an author.

1

0

33

Yen-Huan Li

@yenhuan_li

2 years

More and more people tend to hide the actual connections to previous works when writing papers. Among them, the relatively honest ones reveal the actual connections in the proofs; the others just reproduce essentially the same proofs without any citation.

2

3

32

Yen-Huan Li

@yenhuan_li

4 months

==== My recommendations today ==== In-and-Out: Algorithmic Diffusion for Sampling Convex Bodies A Symplectic Analysis of Alternating Mirror Descent Adaptive Accelerated Composite Minimization (1/n)

2

4

30

Yen-Huan Li

@yenhuan_li

5 months

==== My recommendations today ==== Horoballs and the subgradient method A short proof of the Dvoretzky--Kiefer--Wolfowitz--Massart inequality (1/2)

1

3

31

Yen-Huan Li

@yenhuan_li

2 years

TMLR gives the reviewers two weeks to complete their reviews and basically does not allow the reviewers to reject review assignments. Besides being nice to the authors, I think TMLR should also consider being nicer to the reviewers...

2

1

30

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== Large Stepsize Gradient Descent for Logistic Loss: Non-Monotonicity of the Loss Improves Optimization Efficiency (1/2)

2

6

29

Yen-Huan Li

@yenhuan_li

1 year

We should find some way to encourage researchers to *not skip* words like "the proof here mimics the proof of Theorem A in Paper B", "below is a direct generalization of Lemma C by D et al.", etc.

2

28

Yen-Huan Li

@yenhuan_li

2 years

Learning, information theory, and security seem to be three faces of the same thing. Is there any rigorous statement on their relation?

4

0

28

Yen-Huan Li

@yenhuan_li

4 months

See: - An Explanation for the Logarithmic Connection between Linear and Morphological System Theory - What Shape Is Your Conjugate? A Survey of Computational Convex Analysis and Its Applications (2/2)

What Shape Is Your Conjugate? A Survey of Computational Convex Analysis and Its Applications | SIAM...

Computational convex analysis algorithms have been rediscovered several times in the past by researchers from different fields. To further communications between practitioners, we review the field of...

epubs.siam.org

2

1

27

Yen-Huan Li

@yenhuan_li

3 months

There are too many new papers in arXiv cs.LG every day... We should start distinguishing between (1) papers on ML and (2) papers applying ML tools to their specific research fields. Not all papers involving math equations are submitted to .

2

1

25

Yen-Huan Li

@yenhuan_li

3 months

==== My recommendations today ==== Online conformal prediction with decaying step sizes Dual VC Dimension Obstructs Sample Compression by Embeddings (1/n)

1

3

26

Yen-Huan Li

@yenhuan_li

1 month

A template for presenting research results: 1. What is the problem? 2. Why is it important? 3. Why is it challenging? 7. What are the existing achievements? 5. Why are they unsatisfactory? 6. What is your result? 7. What are the key ideas in the result? 8. Outlook.

2

1

24

Yen-Huan Li

@yenhuan_li

4 years

Winter School on the Interplay between High-Dimensional Geometry and Probability

www.youtube.com

0

7

24

Yen-Huan Li

@yenhuan_li

1 year

What should I do? Recently we found an arXiv paper that generalized our work, which is great. But that arXiv paper simply cited our work as "a special case," hiding the fact that both the algorithm and analysis they consider are directly adopted from our work. (1/n)

5

0

23

Yen-Huan Li

@yenhuan_li

4 months

==== My recommendations today ==== Accelerated Objective Gap and Gradient Norm Convergence for Gradient Descent via Long Steps Optimal Acceleration for Minimax and Fixed-Point Problems is Not Unique (1/n)

2

5

21

Yen-Huan Li

@yenhuan_li

3 years

The reviewer errors section of the reviewer tutorial for ICML 2022 is worth reading. I have experienced at least one example per slide in that section...

ICML 2022 How to be a good reviewer-tutorials for ICML2022 reviewers.pdf

drive.google.com

0

9

21

Yen-Huan Li

@yenhuan_li

3 years

There are two possible reasons for writing a very long paper: 1) the paper is essentially very complicated; 2) the authors haven't fully understood their result. Nowadays, the reviewers tend to think in 1), so the authors tend to stay in 2).

0

1

20

Yen-Huan Li

@yenhuan_li

4 years

Currently this is the only sentence I can think of for rebuttal: dear reviewers, please read the "related work" section again.

0

20

Yen-Huan Li

@yenhuan_li

1 year

My academic idol.

0

18

Yen-Huan Li

@yenhuan_li

2 years

Selected as a NeurIPS top reviewer again.

1

0

17

Yen-Huan Li

@yenhuan_li

5 months

==== My recommendation today ==== Calibration for Decision Making: A Principled Approach to Trustworthy ML

0

1

15

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== An Elementary Predictor Obtaining 2\sqrt{T} Distance to Calibration On Averaging and Extrapolation for Gradient Descent

0

2

15

Yen-Huan Li

@yenhuan_li

3 years

After I spent several hours reviewing a long paper for NuerIPS, I found that the paper disappeared from my list of assigned papers...

3

0

14

Yen-Huan Li

@yenhuan_li

4 years

Kwangjun Ahn: From Proximal Point Method to Nesterov's Acceleration

Understanding Nesterov's Acceleration via Proximal Point Method

The proximal point method (PPM) is a fundamental method in optimization that is often used as a building block for designing optimization algorithms. In this work, we use the PPM method to provide...

arxiv.org

0

12

Yen-Huan Li

@yenhuan_li

4 years

1. Build a review system on arXiv, where experts can share and discuss about their comments about a paper publicly. 2. Reform journals as "recommended lists of arXiv papers" by the editors. 3. Reform conferences as meetings of researchers. 4. Hence no submissions & deadlines.

1

2

13

Yen-Huan Li

@yenhuan_li

5 years

A Modern Introduction to Online Learning. (arXiv:1912.13213v1 [cs.LG]) 來自 @Inoreader

0

1

11

Yen-Huan Li

@yenhuan_li

3 years

The Computer Science Department at National Taiwan University have faculty openings at all ranks beginning in August 2022.

0

1

12

Yen-Huan Li

@yenhuan_li

5 months

Let's eliminate the author-reviewer discussion phase and perhaps replace it with an additional round of author rebuttal. It's insane to expect everyone in academia, spread across various time zones, to be available for real-time discussions over several days.

1

2

11

Yen-Huan Li

@yenhuan_li

3 years

The schedule of the virtual Machine Learning Summer School 2021, Taipei is out!

0

2

11

Yen-Huan Li

@yenhuan_li

4 years

@hanzhao_ml I once got a similar comment and asked the reviewer to provide a reference. The reviewer did not reply and changed the score from five to one.

2

0

11

Yen-Huan Li

@yenhuan_li

4 years

Complexity Theory, Game Theory, and Economics: The Barbados Lectures. By Tim Roughgarden. Free version on arXiv.

0

5

11

Yen-Huan Li

@yenhuan_li

2 years

Difficult to politely remind the reviewers that the issues they raised are all answered in the paper.

0

10

Yen-Huan Li

@yenhuan_li

4 months

I am happy to see such serious issues being discussed publicly.

Christian Wolf

@chriswolfvision

4 months

🤯 #iclr2024

11

46

326

0

10

Yen-Huan Li

@yenhuan_li

2 years

I always encourage my students to pursue "normal" (less exciting?) proofs. If we prove a result by an ad hoc proof, we can spend months trying to find a "normaler" one. I believe doing so is beneficial to other researchers, though I know there are so many disincentives...

Kwang-Sung Jun

@kwangsungjun

2 years

The meta reviewer says "the proof is standard" and rejects a paper that solves an open problem and empirically works better than existing algorithms. Isn't it better if one solves the open problem with standard techniques rather than an ad hoc, very specialized technique?

1

2

59

0

10

Yen-Huan Li

@yenhuan_li

2 years

Our work provides a novel perspective on online portfolio selection and learning quantum states, improving on the regret analysis of $\widetilde{\text{EG}}$ and proposing state-of-the-art algorithms as applications of *one* regret bound.

1

0

9

Yen-Huan Li

@yenhuan_li

1 year

The book I always consult whenever I encounter an unfamiliar concept in statistics.

Reminiscences of a Statistician

link.springer.com

0

1

8

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== Making SGD Parameter-Free The effect of smooth parametrizations on nonconvex optimization landscapes

1

2

9

Yen-Huan Li

@yenhuan_li

5 months

Eh... I also made some careless mistakes in my review comments.

1

0

9

Yen-Huan Li

@yenhuan_li

1 year

Now since there are arXiv and similar preprint archives, we may replace the traditional concept of a journal by regularly announced lists of recommended papers with links. Then, everyone can run their own “journals.” Journals by those with good tastes become the “top” ones.

1

0

9

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== Learnability Gaps of Strategic Classification An optimal tradeoff between entanglement and copy complexity for state tomography

1

9

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== New Brunn--Minkowski and functional inequalities via convexity of entropy Liran Rotem: Brunn-Minkowski inequalities via concavity of entropy (1/2)

2

9

Yen-Huan Li

@yenhuan_li

2 years

Probability in high-dimensions. Lecture notes by Joel Tropp.

0

2

8

Yen-Huan Li

@yenhuan_li

5 months

Excellent lineup of speakers!

Jonathan Scarlett

@j_m_scarlett

5 months

[Reposting with corrected date!] Math of Data Science "Summer" (Australian Winter) School in Darwin June 24-28, with mini-courses from Gabor Lugosi, Peter Bartlett, myself, Subhro Ghosh, Nikita Zhivotovskiy, and Shahar Mendelson.

1

10

41

2

1

9

Yen-Huan Li

@yenhuan_li

5 months

Many theory papers claim to have proved what others cannot, but they often fail to spotlight their key ideas. This confuses reviewers and readers, and hints at a lack of novelty.

1

0

9

Yen-Huan Li

@yenhuan_li

3 months

==== My recommendations today ==== Primal Subgradient Methods with Predefined Step Sizes Sara van de Geer: A statistician's selection of the work of Michel Talagrand (1/n)

1

9

Yen-Huan Li

@yenhuan_li

5 years

"RL, DL, and the role of policy gradient methods" by Sham Kakade via @YouTube

Sham Kakade (U. of Washington) -- The provable effectiveness of...

MIFODS Workshop on Learning with Complex StructureCambridge, USJanuary 27-29, 2020

www.youtube.com

0

1

8

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== Bayesian nonparametric statistics, St-Flour lecture notes Behavior of Sequential Predictors of Binary Sequences (1/2)

3

2

7

Yen-Huan Li

@yenhuan_li

3 months

Scalability is the issue if we want to maintain the current peer review systems. But I would suggest looking for an alternative…

TimDarcet

@TimDarcet

3 months

Current state of neurips abstract submissions This neurips is gonna be crazy

11

22

122

2

0

8

Yen-Huan Li

@yenhuan_li

3 years

Is there any name for the fact that the Fenchel conjugate of the relative entropy is a log-partition function? Not sure if Donsker-Varadhan is appropriate...

3

0

7

Yen-Huan Li

@yenhuan_li

6 months

==== My recommendations today ==== Discussion meeting on ‘Safe Testing’ by Grünwald et al. Merging sequential e-values via martingales Improving Adaptive Online Learning Using Refined Discretization

0

7

Yen-Huan Li

@yenhuan_li

2 months

Generalization bounds for mixing processes via delayed online-to-PAC conversions Improved bounds for calibration via stronger sign preservation games (5/n)

1

5

Yen-Huan Li

@yenhuan_li

4 years

Yurii Nesterov: Superfast Second-Order Methods for Unconstrained Convex Optimization @ One World Optimization Seminar via @YouTube

OWOS: Yurii Nesterov - "Superfast Second-Order Methods for Unconstr...

The eleventh talk in the One World Optimization Seminar given on June 29th, 2020, by Prof. Yurii Nesterov (University of Louvain) on "Superfast Second-Order ...

www.youtube.com

0

7

Yen-Huan Li

@yenhuan_li

5 months

From Complexity to Clarity: Analytical Expressions of Deep Neural Network Weights via Clifford's Geometric Algebra and Convexity (2/2)

0

7

Yen-Huan Li

@yenhuan_li

5 months

@miniapeur Kolmogorov, von Neumann, Shannon.

0

7

Yen-Huan Li

@yenhuan_li

2 years

"However, even after the rebuttal it remained unclear as to why XXX is fruitful, and whether it leads to theoretical or empirical improvement over prior works." After rebuttal it remained also unclear to me whether anyone read the rebuttal...

2

0

7

Yen-Huan Li

@yenhuan_li

1 year

@gabrielpeyre A related very new paper:

A Hamilton–Jacobi-based proximal operator | PNAS

First-order optimization algorithms are widely used today. Two standard building blocks in these algorithms are proximal operators (proximals) and ...

www.pnas.org

0

7

Yen-Huan Li

@yenhuan_li

4 years

This is the first time almost all of the reviewers and ACs of the papers I reviewed for NeurIPS are decent. Very surprising. Unfortunately our submission did not encounter similar reviewers.

0

7

Yen-Huan Li

@yenhuan_li

4 months

It seems that practitioners and theoreticians have different opinions on this matter. Perhaps the true issue lies in enforcing one single system, one uniform rule for all. Maintaining the diversity & openness of the ML community has been always challenging.

Gautam Kamath

@thegautamkamath

4 months

NeurIPS 2024 will have a track for papers from high schoolers.

79

90

598

0

1

7

Yen-Huan Li

@yenhuan_li

2 years

Looks very interesting. Contains some articles on the history of probability theory. Book: The Splendors and Miseries of Martingales

The Splendors and Miseries of Martingales

link.springer.com

0

6

Yen-Huan Li

@yenhuan_li

4 years

Faculty position opening --- Department of Computer Science and Information Engin eering at National Taiwan University

0

1

7

Yen-Huan Li

@yenhuan_li

1 month

Open problem: Direct Sums in Learning Theory Calibrated Forecasts: The Minimax Proof Analysis of Langevin Monte Carlo from Poincaré to Log-Sobolev (2/n)

1

2

6