TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Fast Randomized SVD (2014)

54 pointsby Cynddlabout 10 years ago

3 comments

Cynddlabout 10 years ago
Posted six months ago on <a href="https://news.ycombinator.com/item?id=8525237" rel="nofollow">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=8525237</a>.<p>&gt; We will soon release the implementations for these algorithms described.<p>I would like to see them now.
评论 #9264406 未加载
评论 #9264401 未加载
rwittenabout 10 years ago
If you&#x27;re interested in lower bounds or tighter upper bounds, you can find the latest here:<p><a href="http://link.springer.com/article/10.1007%2Fs00453-014-9891-7#page-1" rel="nofollow">http:&#x2F;&#x2F;link.springer.com&#x2F;article&#x2F;10.1007%2Fs00453-014-9891-7...</a> <a href="http://statweb.stanford.edu/~candes/papers/RandomizedNLA.pdf" rel="nofollow">http:&#x2F;&#x2F;statweb.stanford.edu&#x2F;~candes&#x2F;papers&#x2F;RandomizedNLA.pdf</a>
inglorabout 10 years ago
Why aren&#x27;t they just using compressed sensing instead of PCA in the first place? PCA is good because it guarantees perfect recovery wen the set of examples is contained in an n dimentional subspace. Compressed sensing guarantees recovery whenever the set of examples is sparse in <i>some basis</i> - it sounds like a much better fit.<p>Not to mention random projections which are even faster (even proved by the Johnson-Lindenstrauss lemma) usually do well,
评论 #9267736 未加载
评论 #9265094 未加载