Accepted (and not accepted) papers for @iclr_conf are now visible and de-anonymized on OpenReview #ICLR2023 https://
openreview.net/group?id=ICLR.
cc/2023/Conference
…
@thegautamkamath
-
ICLR 2023 Accepted Papers Now Visible on OpenReview
By
–
-
AI Tools Disclosure Standards for Academic Authors
By
–
1. Authors should report tools they use (consistent with field standards)
2. Authors always take responsibility for paper contents
3. Generative AI should not be listed as an author Seems sensible enough. -
Trade-off Optimization Using Higher Order Distribution Moments
By
–
Yes! The trade-off gets better if you have more moments for the underlying distribution (versus just a variance). We have the theorem in the paper.
-
Research Impact Beyond Papers: Building Cumulative Understanding
By
–
Many have the misconception that research is just about publishing papers. In my view, research is about creating a body of work that, when taken in aggregate, extends our understanding. Papers are just stepping stones. This seems like a nice initiative to refocus things.
-
Unbiased Algorithms in Differential Privacy Research
By
–
BONUS TWEET Be sure to also check out this other paper posted at the same time, by @thesasho and Haohua Tang, also focused on unbiased algorithms in differential privacy. Despite similarities in the titles, the settings are mostly different. https://
arxiv.org/abs/2301.13850 9/8 -
Congratulations to Junior Researchers on Successful AI Paper
By
–
Anyway, this was a very satisfying paper (
https://
x.com/shortstein/sta
tus/1620279198333149184
…), and I think we really solved everything we set out to. Congrats to junior researchers @argymouz Matthew Regehr @vkerdos on this nice paper! https://
arxiv.org/abs/2301.13334 8/8 -
Differential Privacy Limitations for Gaussian Distribution Estimation
By
–
But we also show that pure (epsilon, 0)-DP is hopeless, even for really simple classes like Gaussians, and that the delta in approx DP is needed to perform unbiased estimation. 7/n
-
Unbiased Estimators for Symmetric Distributions
By
–
So that's sad. Is there any hope for special cases? If we happen to know the underlying distribution is symmetric, then *yes*, we can get unbiased estimators. 6/n
-
Privacy-Bias-Variance Trilemma in Mean Estimators
By
–
Our main result: no. There is a *trilemma* between privacy, bias, and variance of a mean estimator: essentially, one can not simultaneously have strong privacy, low bias, and low variance. This shows the clip-and-noise algorithm is optimal. 5/n
-
Unbiased Estimators: Better Algorithms for ML
By
–
This can be undesirable in a number of settings, when unbiased estimators are preferred. E.g., then we can compute the statistic multiple times on independent datasets and average them to reduce error. Natural question: are there better algorithms with no (or low) bias? 4/n