Top all-time · #8208
Francesco Orabona
Personal blog · parameterfree.comaimath
Francesco Orabona writes Parameter-free Learning and Optimization Algorithms, a personal blog at parameterfree.com about ai and math. Since 2020, 10 different Hacker News readers have linked it 13 times, most often “An ICML Award Fiasco”. It is still writing: the last post was on 2026-05-02, and you can follow it by feed.
Too few sampled comments (3) to say how often HN recommends it.?
Why HN links it
HN readers link Francesco Orabona's blog for its machine learning posts on optimizers and training, such as schedule-free learning, Adam, and theoretical limits of transformers. The comments cite these posts without judging them.
Summarised by Claude from a sample of 3 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Because life is too short to tune learning rates. By Francesco Orabona
Most-linked posts?
- An ICML Award Fiasco5 people · 5 mentions · first linked 2023-08-30
- Neural Networks (Maybe) Evolved to Make Adam the Best Optimizer3 people · 5 mentions · first linked 2020-12-07
- The Myth of the Expert Reviewer2 people · 2 mentions · first linked 2021-07-06
What Hacker News says?
I beg to disagree. [1] provides one with a whole-data-set training method (ADMM, one of such methods). Page 8 contains figure 2(b) - accuracy of training after specified amount of time. Note that ADMM start where stochastic gradient stops. [1] https://arxiv.org/pdf/1605.02026.pdf At [2] I tried to apply logistic regression trained using reweighted least squares algorithm on the same Higgs boson data set. I've got th…
Comment linking /2020/12/06/neural-network-maybe-evolved-to-make-adam-the-best-optimizer/
on “Google Open-Sources Trillion-Parameter AI Language Model Switch Transformer”Submitted An ICML Award Fiasco as a story · 2 points
HN discussionI was asking on Twitter if Aaron had any experiments for transformers, since they provided some graphs for CNNs and the like, but no transformers. * Aaron et al's past work on D-Adaptation won a best ICML paper, with their follow up work being Prodigy - but both on transformers did similar or worse than AdamW. https://twitter.com/danielhanchen/status/1775547139248341125 * Superconvergence + LR range finder + Fast AI…
Comment linking /2023/08/30/yet-another-icml-award-fiasco/
on “Schedule-Free Learning – A New Way to Train”Submitted The Myth of the Expert Reviewer as a story · 2 points
HN discussionJust like with neural networks and Adam [1], LLMs evolve to make transformers their best building block. [1] https://parameterfree.com/2020/12/06/neural-network-maybe-ev...
Comment linking /2020/12/06/neural-network-maybe-evolved-to-make-adam-the-best-optimizer/
on “Theoretical limitations of multi-layer Transformer”Submitted A Minimizer Far, Far Away as a story · 1 points
HN discussionSubmitted Neural Networks (Maybe) Evolved to Make Adam the Best Optimizer as a story · 1 points
HN discussionSubmitted Yet Another ICML Award Fiasco as a story · 1 points
HN discussionSubmitted Yet another ICML award fiasco as a story · 1 points
HN discussionSubmitted Yet Another ICML Award Fiasco as a story · 1 points
HN discussionSubmitted The Myth of the Expert Reviewer as a story · 1 points
HN discussionSubmitted Neural Networks (Maybe) Evolved to Make Adam the Best Optimizer as a story · 1 points
HN discussionSubmitted Neural Networks (Maybe) Evolved to Make Adam the Best Optimizer as a story · 1 points
HN discussion
