Matches in SemOpenAlex for { <https://semopenalex.org/work/W3100743822> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W3100743822 endingPage "5428" @default.
- W3100743822 startingPage "5417" @default.
- W3100743822 abstract "We consider the problem of learning the best-fitting single neuron as measured by the expected square loss $mathbb{E}_{(x,y)sim mathcal{D}}[(sigma(w^top x)-y)^2]$ over some unknown joint distribution $mathcal{D}$ by using gradient descent to minimize the empirical risk induced by a set of i.i.d. samples $Ssim mathcal{D}^n$. The activation function $sigma$ is an arbitrary Lipschitz and non-decreasing function, making the optimization problem nonconvex and nonsmooth in general, and covers typical neural network activation functions and inverse link functions in the generalized linear model setting. In the agnostic PAC learning setting, where no assumption on the relationship between the labels $y$ and the input $x$ is made, if the optimal population risk is $mathsf{OPT}$, we show that gradient descent achieves population risk $O(mathsf{OPT})+epsilon$ in polynomial time and sample complexity when $sigma$ is strictly increasing. For the ReLU activation, our population risk guarantee is $O(mathsf{OPT}^{1/2})+epsilon$. When labels take the form $y = sigma(v^top x) + xi$ for zero-mean sub-Gaussian noise $xi$, we show that the population risk guarantees for gradient descent improve to $mathsf{OPT} + epsilon$. Our sample complexity and runtime guarantees are (almost) dimension independent, and when $sigma$ is strictly increasing, require no distributional assumptions beyond boundedness. For ReLU, we show the same results under a nondegeneracy assumption for the marginal distribution of the input." @default.
- W3100743822 created "2020-11-23" @default.
- W3100743822 creator A5035797060 @default.
- W3100743822 creator A5051448391 @default.
- W3100743822 creator A5086329948 @default.
- W3100743822 date "2020-05-29" @default.
- W3100743822 modified "2023-09-24" @default.
- W3100743822 title "Agnostic learning of a single neuron with gradient descent" @default.
- W3100743822 hasPublicationYear "2020" @default.
- W3100743822 type Work @default.
- W3100743822 sameAs 3100743822 @default.
- W3100743822 citedByCount "2" @default.
- W3100743822 countsByYear W31007438222020 @default.
- W3100743822 countsByYear W31007438222021 @default.
- W3100743822 crossrefType "proceedings-article" @default.
- W3100743822 hasAuthorship W3100743822A5035797060 @default.
- W3100743822 hasAuthorship W3100743822A5051448391 @default.
- W3100743822 hasAuthorship W3100743822A5086329948 @default.
- W3100743822 hasConcept C110121322 @default.
- W3100743822 hasConcept C114614502 @default.
- W3100743822 hasConcept C118615104 @default.
- W3100743822 hasConcept C121332964 @default.
- W3100743822 hasConcept C134306372 @default.
- W3100743822 hasConcept C14036430 @default.
- W3100743822 hasConcept C144024400 @default.
- W3100743822 hasConcept C149923435 @default.
- W3100743822 hasConcept C153258448 @default.
- W3100743822 hasConcept C154945302 @default.
- W3100743822 hasConcept C207467116 @default.
- W3100743822 hasConcept C22324862 @default.
- W3100743822 hasConcept C2524010 @default.
- W3100743822 hasConcept C2778049214 @default.
- W3100743822 hasConcept C2908647359 @default.
- W3100743822 hasConcept C33676613 @default.
- W3100743822 hasConcept C33923547 @default.
- W3100743822 hasConcept C41008148 @default.
- W3100743822 hasConcept C50644808 @default.
- W3100743822 hasConcept C62520636 @default.
- W3100743822 hasConcept C78458016 @default.
- W3100743822 hasConcept C86803240 @default.
- W3100743822 hasConcept C90119067 @default.
- W3100743822 hasConceptScore W3100743822C110121322 @default.
- W3100743822 hasConceptScore W3100743822C114614502 @default.
- W3100743822 hasConceptScore W3100743822C118615104 @default.
- W3100743822 hasConceptScore W3100743822C121332964 @default.
- W3100743822 hasConceptScore W3100743822C134306372 @default.
- W3100743822 hasConceptScore W3100743822C14036430 @default.
- W3100743822 hasConceptScore W3100743822C144024400 @default.
- W3100743822 hasConceptScore W3100743822C149923435 @default.
- W3100743822 hasConceptScore W3100743822C153258448 @default.
- W3100743822 hasConceptScore W3100743822C154945302 @default.
- W3100743822 hasConceptScore W3100743822C207467116 @default.
- W3100743822 hasConceptScore W3100743822C22324862 @default.
- W3100743822 hasConceptScore W3100743822C2524010 @default.
- W3100743822 hasConceptScore W3100743822C2778049214 @default.
- W3100743822 hasConceptScore W3100743822C2908647359 @default.
- W3100743822 hasConceptScore W3100743822C33676613 @default.
- W3100743822 hasConceptScore W3100743822C33923547 @default.
- W3100743822 hasConceptScore W3100743822C41008148 @default.
- W3100743822 hasConceptScore W3100743822C50644808 @default.
- W3100743822 hasConceptScore W3100743822C62520636 @default.
- W3100743822 hasConceptScore W3100743822C78458016 @default.
- W3100743822 hasConceptScore W3100743822C86803240 @default.
- W3100743822 hasConceptScore W3100743822C90119067 @default.
- W3100743822 hasLocation W31007438221 @default.
- W3100743822 hasOpenAccess W3100743822 @default.
- W3100743822 hasPrimaryLocation W31007438221 @default.
- W3100743822 hasRelatedWork W1582126688 @default.
- W3100743822 hasRelatedWork W2613688792 @default.
- W3100743822 hasRelatedWork W2765733029 @default.
- W3100743822 hasRelatedWork W2767674840 @default.
- W3100743822 hasRelatedWork W2770027754 @default.
- W3100743822 hasRelatedWork W2791562219 @default.
- W3100743822 hasRelatedWork W2791759757 @default.
- W3100743822 hasRelatedWork W2804066051 @default.
- W3100743822 hasRelatedWork W2901741727 @default.
- W3100743822 hasRelatedWork W2938288422 @default.
- W3100743822 hasRelatedWork W2964271537 @default.
- W3100743822 hasRelatedWork W2966991718 @default.
- W3100743822 hasRelatedWork W2970890756 @default.
- W3100743822 hasRelatedWork W2986078788 @default.
- W3100743822 hasRelatedWork W3035067130 @default.
- W3100743822 hasRelatedWork W3046665994 @default.
- W3100743822 hasRelatedWork W3088153346 @default.
- W3100743822 hasRelatedWork W3100482347 @default.
- W3100743822 hasRelatedWork W3109699007 @default.
- W3100743822 hasRelatedWork W3175284970 @default.
- W3100743822 hasVolume "33" @default.
- W3100743822 isParatext "false" @default.
- W3100743822 isRetracted "false" @default.
- W3100743822 magId "3100743822" @default.
- W3100743822 workType "article" @default.