Is LogSumExp Optimal? A Peer-Reviewed Answer in the Journal of Complexity
prof_grimmer · x · 2026-11-18
profgrimmer and student Thabo Samakhoana investigated softmax/LogSumExp smoothing — the de facto standard across ML and optimization — asking in what sense the choice is "optimal". They found elementary answers, with both good and bad news, published in the Journal of Complexity. The author praises peer review for challenging and maturing the work, and shares how the paper grew from its initial version.
More from Research
- In LLM quant research, eval design beats prompt engineering — bgoncalves · 2026-11-18
- Blink CPU speed test: model2vec 3.8x faster solo, Sentence Transformers 2.2x faster at batch 1024 — tomaarsen · 2026-11-18
- blink-8M tops seven static embedding models with 55.35 MTEB and 46.70 BEIR nDCG@10 — tomaarsen · 2026-11-18
- blink-multilingual-64M demo: English query retrieves Dutch and Spanish docs correctly — tomaarsen · 2026-11-18
- MinishLab ships Blink static embedding models: 8M and 64M params, fast CPU retrieval — tomaarsen · 2026-11-18
- Massive proof-dependency graph for Theorem 3.1.1 wows math Twitter as formalization scales up — littmath · 2026-11-18