Understanding Short-Horizon Bias in Stochastic Meta-Optimization · HackerPapers discussionshttps://hackerpaper.kaipfister.de/p/1803.02021/Collected HN comments and submissions linking this paper.enMon, 02 Nov 2026 11:16:22 GMTUnderstanding Short-Horizon Bias in Stochastic Meta-Optimization — mentioned by Strawhttps://news.ycombinator.com/item?id=21145219urn:hackerpapers:mention:1803.02021:21145219Thu, 03 Oct 2019 09:50:34 GMT<p>Unfortunately, the 1-step optimal learning rate often differs massively from the long-horizon best choice: https://arxiv.org/abs/1803.02021 Due to the fact that larger LRs can result in worse immediate performance but better progress in low signal-to-noise ratio directions.</p><p><a href="https://hackerpaper.kaipfister.de/p/1803.02021/">Paper and discussion history</a>. A mention is not an endorsement.</p>cs.LG