FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3100482347> ?p ?o ?g. }

• W3100482347 abstract "Stochastic gradient descent (SGD) is a popular algorithm for optimization problems arising in high-dimensional inference tasks. Here one produces an estimator of an unknown parameter from independent samples of data by iteratively optimizing a loss function. This loss function is random and often non-convex. We study the performance of the simplest version of SGD, namely online SGD, from a random start in the setting where the parameter space is high-dimensional. We develop nearly sharp thresholds for the number of samples needed for consistent estimation as one varies the dimension. Our thresholds depend only on an intrinsic property of the population loss which we call the information exponent. In particular, our results do not assume uniform control on the loss itself, such as convexity or uniform derivative bounds. The thresholds we obtain are polynomial in the dimension and the precise exponent depends explicitly on the information exponent. As a consequence of our results, we find that except for the simplest tasks, almost all of the data is used simply in the initial search phase to obtain non-trivial correlation with the ground truth. Upon attaining non-trivial correlation, the descent is rapid and exhibits law of large numbers type behavior. We illustrate our approach by applying it to a wide set of inference tasks such as phase retrieval, and parameter estimation for generalized linear models, online PCA, and spiked tensor models, as well as to supervised learning for single-layer networks with general activation functions." @default.
• W3100482347 created "2020-11-23" @default.
• W3100482347 creator A5039747459 @default.
• W3100482347 creator A5051194866 @default.
• W3100482347 creator A5085479654 @default.
• W3100482347 date "2020-03-23" @default.
• W3100482347 modified "2023-09-27" @default.
• W3100482347 title "Online stochastic gradient descent on non-convex losses from high-dimensional inference" @default.
• W3100482347 cites W1491706803 @default.
• W3100482347 cites W1520752838 @default.
• W3100482347 cites W1528905581 @default.
• W3100482347 cites W1568229137 @default.
• W3100482347 cites W1644055776 @default.
• W3100482347 cites W1663973292 @default.
• W3100482347 cites W188867022 @default.
• W3100482347 cites W1891022888 @default.
• W3100482347 cites W1994616650 @default.
• W3100482347 cites W2012828271 @default.
• W3100482347 cites W2050583479 @default.
• W3100482347 cites W2063588877 @default.
• W3100482347 cites W2083459869 @default.
• W3100482347 cites W2090980101 @default.
• W3100482347 cites W2102486516 @default.
• W3100482347 cites W2112796928 @default.
• W3100482347 cites W2125812768 @default.
• W3100482347 cites W2127300249 @default.
• W3100482347 cites W2132211083 @default.
• W3100482347 cites W2159062998 @default.
• W3100482347 cites W2199563553 @default.
• W3100482347 cites W2290522125 @default.
• W3100482347 cites W2313324941 @default.
• W3100482347 cites W2399198888 @default.
• W3100482347 cites W2552702319 @default.
• W3100482347 cites W2557283755 @default.
• W3100482347 cites W2589022458 @default.
• W3100482347 cites W2590513847 @default.
• W3100482347 cites W2606981059 @default.
• W3100482347 cites W2725016515 @default.
• W3100482347 cites W2772785876 @default.
• W3100482347 cites W2802698595 @default.
• W3100482347 cites W2891786606 @default.
• W3100482347 cites W2899900971 @default.
• W3100482347 cites W2918745211 @default.
• W3100482347 cites W2955330387 @default.
• W3100482347 cites W2962915600 @default.
• W3100482347 cites W2963060833 @default.
• W3100482347 cites W2963122491 @default.
• W3100482347 cites W2963244042 @default.
• W3100482347 cites W2963595633 @default.
• W3100482347 cites W2963811132 @default.
• W3100482347 cites W2963877580 @default.
• W3100482347 cites W2964038363 @default.
• W3100482347 cites W2964106499 @default.
• W3100482347 cites W2965944281 @default.
• W3100482347 cites W2968353065 @default.
• W3100482347 cites W2970759752 @default.
• W3100482347 cites W2972186810 @default.
• W3100482347 cites W2978927916 @default.
• W3100482347 cites W3023124831 @default.
• W3100482347 cites W3034877436 @default.
• W3100482347 cites W3043072433 @default.
• W3100482347 cites W3044069306 @default.
• W3100482347 cites W3046968338 @default.
• W3100482347 cites W3101714678 @default.
• W3100482347 cites W3101975788 @default.
• W3100482347 cites W3102677403 @default.
• W3100482347 cites W3103328814 @default.
• W3100482347 cites W3105254325 @default.
• W3100482347 cites W3113425034 @default.
• W3100482347 cites W3114970390 @default.
• W3100482347 cites W3116143161 @default.
• W3100482347 cites W3123272904 @default.
• W3100482347 cites W653925602 @default.
• W3100482347 hasPublicationYear "2020" @default.
• W3100482347 type Work @default.
• W3100482347 sameAs 3100482347 @default.
• W3100482347 citedByCount "0" @default.
• W3100482347 crossrefType "posted-content" @default.
• W3100482347 hasAuthorship W3100482347A5039747459 @default.
• W3100482347 hasAuthorship W3100482347A5051194866 @default.
• W3100482347 hasAuthorship W3100482347A5085479654 @default.
• W3100482347 hasConcept C105795698 @default.
• W3100482347 hasConcept C106159729 @default.
• W3100482347 hasConcept C112680207 @default.
• W3100482347 hasConcept C11413529 @default.
• W3100482347 hasConcept C114614502 @default.
• W3100482347 hasConcept C126255220 @default.
• W3100482347 hasConcept C138885662 @default.
• W3100482347 hasConcept C14036430 @default.
• W3100482347 hasConcept C153258448 @default.
• W3100482347 hasConcept C154945302 @default.
• W3100482347 hasConcept C157972887 @default.
• W3100482347 hasConcept C162324750 @default.
• W3100482347 hasConcept C185429906 @default.
• W3100482347 hasConcept C206688291 @default.
• W3100482347 hasConcept C2524010 @default.
• W3100482347 hasConcept C2776214188 @default.
• W3100482347 hasConcept C2780388253 @default.
• W3100482347 hasConcept C28826006 @default.
• W3100482347 hasConcept C33676613 @default.

• next