Why does L1 regularization induce sparse models?. Many illustrate this using the least squares problem with a norm constraint. The least squares level sets are drawn next to the different unit "circles". I prepared a cool animation which I believe makes it even clearer 🙂

RT @charles0neill: Today, we’re launching Parsed. We are incredibly lucky to live in a world where we stand on the shoulders of giants, fir….

RT @gail_w: Reviewing for neurips has been not even thankless labour this year, but it’s also yielded this incredible email, so who’s reall….

RT @bremen79: Dear Reviewers,. It is completely fine to admit you were wrong in your initial evaluations. You will not lose anything, and t….

Today I had to explain to a reviewer the difference between proving a reduction and proving an equivalence (of two algorithms). 🥲.

When authors thank all other "helpful reviews" but then call your review "insightful":

Is it appropriate for a reviewer to comment on another review, or on the authors' response to that review?. Especially if it seems the authors may be (perhaps unintentionally) misleading the other reviewer?. #NeurIPS2025.

That's also how I detect many LLM generated reviews. They're well written, but say nothing helpful. They don't point out missing comparisons/references. They don't find flaws in the logical flow. They just turn the abstract to "strengths" and the future work to "weaknesses".

A good review should generally *add* information to the process. If the review only summarizes what the authors already wrote in their abstract and intro+limitation sections, it's not really a review but rather a summary. And for that, we don't need humans anymore. #NeurIPS2025.

I think an LLM would have actually done a *better* job than some of the #NeurIPS2025 reviewers this year.

RT @ffabffrasca: “Balancing Efficiency and Expressiveness: Subgraph GNNs with Walk-Based Centrality” is at #ICML2025!. Drop by our poster o….

RT @miniapeur:

RT @lorenzo_noci: Pass by if you want to know about scaling up your model under distribution shifts of the training data. Take away: muP….

RT @boknilev: With the continuing increase paper submissions to ML conferences, how should the community act?. 1. Scale up: more reviewers,….

RT @ns123abc: META Chief AI officer dropped the official list of everyone META hired for their superintelligence team: .

RT @konstmish: I believe successful neural network training represents cases of "near convexity": the optimization landscape, while technic….

RT @docmilanfar: Iranians, check on your Israeli friends and colleagues. Israelis, check on your Iranian friends and colleagues. Becaus….

RT @VerminusM: For the love of God, do your homework!

RT @jonsrosenfeld: We are hiring postdocs at @MIT_CSAIL !. We are at a high stakes moment in time. If you are passionate about uncovering,….

RT @YoniSlutzky: Do neural nets really need gradient descent to generalize?🚨. We dive into matrix factorization and find a sharp split: wid….

RT @HazanPrinceton: Second paper: .What is better than spectral filtering? Double spectral filtering :), for control of partially observed….