Matches in SemOpenAlex for { <https://semopenalex.org/work/W3165153646> ?p ?o ?g. }
- W3165153646 abstract "We establish upper bounds for the expected excess risk of models trained by proper iterative algorithms which approximate the local minima. Unlike the results built upon the strong globally strongly convexity or global growth conditions e.g., PL-inequality, we only require the population risk to be emph{locally} strongly convex around its local minima. Concretely, our bound under convex problems is of order $tilde{cO}(1/n)$. For non-convex problems with $d$ model parameters such that $d/n$ is smaller than a threshold independent of $n$, the order of $tilde{cO}(1/n)$ can be maintained if the empirical risk has no spurious local minima with high probability. Moreover, the bound for non-convex problem becomes $tilde{cO}(1/sqrt{n})$ without such assumption. Our results are derived via algorithmic stability and characterization of the empirical risk's landscape. Compared with the existing algorithmic stability based results, our bounds are dimensional insensitive and without restrictions on the algorithm's implementation, learning rate, and the number of iterations. Our bounds underscore that with locally strongly convex population risk, the models trained by any proper iterative algorithm can generalize well, even for non-convex problems, and $d$ is large." @default.
- W3165153646 created "2021-06-07" @default.
- W3165153646 creator A5026709102 @default.
- W3165153646 creator A5075715353 @default.
- W3165153646 creator A5087695030 @default.
- W3165153646 date "2020-12-04" @default.
- W3165153646 modified "2023-09-25" @default.
- W3165153646 title "Characterization of Excess Risk for Locally Strongly Convex Population Risk" @default.
- W3165153646 cites W1516903196 @default.
- W3165153646 cites W1574111330 @default.
- W3165153646 cites W1970856737 @default.
- W3165153646 cites W1994616650 @default.
- W3165153646 cites W2018672358 @default.
- W3165153646 cites W2101234009 @default.
- W3165153646 cites W2105875671 @default.
- W3165153646 cites W2107438106 @default.
- W3165153646 cites W2108136473 @default.
- W3165153646 cites W2108475251 @default.
- W3165153646 cites W2111377143 @default.
- W3165153646 cites W2112796928 @default.
- W3165153646 cites W2139338362 @default.
- W3165153646 cites W2154952480 @default.
- W3165153646 cites W2164096571 @default.
- W3165153646 cites W2169502938 @default.
- W3165153646 cites W2194775991 @default.
- W3165153646 cites W2338469198 @default.
- W3165153646 cites W2579923771 @default.
- W3165153646 cites W2619167391 @default.
- W3165153646 cites W2753896574 @default.
- W3165153646 cites W2795605442 @default.
- W3165153646 cites W2807299122 @default.
- W3165153646 cites W2890842400 @default.
- W3165153646 cites W2891057995 @default.
- W3165153646 cites W2899748887 @default.
- W3165153646 cites W2911876570 @default.
- W3165153646 cites W2912260645 @default.
- W3165153646 cites W2951396696 @default.
- W3165153646 cites W2962696654 @default.
- W3165153646 cites W2962698540 @default.
- W3165153646 cites W2962781506 @default.
- W3165153646 cites W2963000090 @default.
- W3165153646 cites W2963013132 @default.
- W3165153646 cites W2963092340 @default.
- W3165153646 cites W2963094221 @default.
- W3165153646 cites W2963122491 @default.
- W3165153646 cites W2963153443 @default.
- W3165153646 cites W2963156201 @default.
- W3165153646 cites W2963248893 @default.
- W3165153646 cites W2963326510 @default.
- W3165153646 cites W2963334018 @default.
- W3165153646 cites W2963349772 @default.
- W3165153646 cites W2963403868 @default.
- W3165153646 cites W2963411541 @default.
- W3165153646 cites W2963433607 @default.
- W3165153646 cites W2963446085 @default.
- W3165153646 cites W2963470657 @default.
- W3165153646 cites W2963563140 @default.
- W3165153646 cites W2963655672 @default.
- W3165153646 cites W2963739978 @default.
- W3165153646 cites W2963794891 @default.
- W3165153646 cites W2963862692 @default.
- W3165153646 cites W2963874210 @default.
- W3165153646 cites W2963956929 @default.
- W3165153646 cites W2964106499 @default.
- W3165153646 cites W2964121744 @default.
- W3165153646 cites W2964156132 @default.
- W3165153646 cites W2965157832 @default.
- W3165153646 cites W2966465943 @default.
- W3165153646 cites W2970498991 @default.
- W3165153646 cites W2972203720 @default.
- W3165153646 cites W3003651450 @default.
- W3165153646 cites W3046489903 @default.
- W3165153646 cites W3046508829 @default.
- W3165153646 cites W3097931149 @default.
- W3165153646 cites W3100231902 @default.
- W3165153646 cites W3118608800 @default.
- W3165153646 cites W3136879969 @default.
- W3165153646 cites W577198184 @default.
- W3165153646 cites W88685657 @default.
- W3165153646 doi "https://doi.org/10.48550/arxiv.2012.02456" @default.
- W3165153646 hasPublicationYear "2020" @default.
- W3165153646 type Work @default.
- W3165153646 sameAs 3165153646 @default.
- W3165153646 citedByCount "0" @default.
- W3165153646 crossrefType "posted-content" @default.
- W3165153646 hasAuthorship W3165153646A5026709102 @default.
- W3165153646 hasAuthorship W3165153646A5075715353 @default.
- W3165153646 hasAuthorship W3165153646A5087695030 @default.
- W3165153646 hasBestOaLocation W31651536461 @default.
- W3165153646 hasConcept C10138342 @default.
- W3165153646 hasConcept C105795698 @default.
- W3165153646 hasConcept C106159729 @default.
- W3165153646 hasConcept C112680207 @default.
- W3165153646 hasConcept C112972136 @default.
- W3165153646 hasConcept C114614502 @default.
- W3165153646 hasConcept C119857082 @default.
- W3165153646 hasConcept C120665830 @default.
- W3165153646 hasConcept C121332964 @default.
- W3165153646 hasConcept C126255220 @default.
- W3165153646 hasConcept C134306372 @default.