Matches in SemOpenAlex for { <https://semopenalex.org/work/W4307531624> ?p ?o ?g. }
Showing items 1 to 62 of
62
with 100 items per page.
- W4307531624 abstract "Polynomial expansions are important in the analysis of neural network nonlinearities. They have been applied thereto addressing well-known difficulties in verification, explainability, and security. Existing approaches span classical Taylor and Chebyshev methods, asymptotics, and many numerical approaches. We find that while these individually have useful properties such as exact error formulas, adjustable domain, and robustness to undefined derivatives, there are no approaches that provide a consistent method yielding an expansion with all these properties. To address this, we develop an analytically modified integral transform expansion (AMITE), a novel expansion via integral transforms modified using derived criteria for convergence. We show the general expansion and then demonstrate application for two popular activation functions, hyperbolic tangent and rectified linear units. Compared with existing expansions (i.e., Chebyshev, Taylor, and numerical) employed to this end, AMITE is the first to provide six previously mutually exclusive desired expansion properties such as exact formulas for the coefficients and exact expansion errors (Table II). We demonstrate the effectiveness of AMITE in two case studies. First, a multivariate polynomial form is efficiently extracted from a single hidden layer black-box Multi-Layer Perceptron (MLP) to facilitate equivalence testing from noisy stimulus-response pairs. Second, a variety of Feed-Forward Neural Network (FFNN) architectures having between 3 and 7 layers are range bounded using Taylor models improved by the AMITE polynomials and error formulas. AMITE presents a new dimension of expansion methods suitable for analysis/approximation of nonlinearities in neural networks, opening new directions and opportunities for the theoretical analysis and systematic testing of neural networks." @default.
- W4307531624 created "2022-11-02" @default.
- W4307531624 creator A5007292377 @default.
- W4307531624 creator A5016146461 @default.
- W4307531624 creator A5018691885 @default.
- W4307531624 date "2020-07-13" @default.
- W4307531624 modified "2023-10-17" @default.
- W4307531624 title "AMITE: A Novel Polynomial Expansion for Analyzing Neural Network Nonlinearities" @default.
- W4307531624 doi "https://doi.org/10.48550/arxiv.2007.06226" @default.
- W4307531624 hasPublicationYear "2020" @default.
- W4307531624 type Work @default.
- W4307531624 citedByCount "0" @default.
- W4307531624 crossrefType "posted-content" @default.
- W4307531624 hasAuthorship W4307531624A5007292377 @default.
- W4307531624 hasAuthorship W4307531624A5016146461 @default.
- W4307531624 hasAuthorship W4307531624A5018691885 @default.
- W4307531624 hasBestOaLocation W43075316241 @default.
- W4307531624 hasConcept C11413529 @default.
- W4307531624 hasConcept C126255220 @default.
- W4307531624 hasConcept C129785596 @default.
- W4307531624 hasConcept C134306372 @default.
- W4307531624 hasConcept C154945302 @default.
- W4307531624 hasConcept C158946198 @default.
- W4307531624 hasConcept C171326582 @default.
- W4307531624 hasConcept C186080144 @default.
- W4307531624 hasConcept C21424316 @default.
- W4307531624 hasConcept C28826006 @default.
- W4307531624 hasConcept C33923547 @default.
- W4307531624 hasConcept C41008148 @default.
- W4307531624 hasConcept C50644808 @default.
- W4307531624 hasConcept C90119067 @default.
- W4307531624 hasConceptScore W4307531624C11413529 @default.
- W4307531624 hasConceptScore W4307531624C126255220 @default.
- W4307531624 hasConceptScore W4307531624C129785596 @default.
- W4307531624 hasConceptScore W4307531624C134306372 @default.
- W4307531624 hasConceptScore W4307531624C154945302 @default.
- W4307531624 hasConceptScore W4307531624C158946198 @default.
- W4307531624 hasConceptScore W4307531624C171326582 @default.
- W4307531624 hasConceptScore W4307531624C186080144 @default.
- W4307531624 hasConceptScore W4307531624C21424316 @default.
- W4307531624 hasConceptScore W4307531624C28826006 @default.
- W4307531624 hasConceptScore W4307531624C33923547 @default.
- W4307531624 hasConceptScore W4307531624C41008148 @default.
- W4307531624 hasConceptScore W4307531624C50644808 @default.
- W4307531624 hasConceptScore W4307531624C90119067 @default.
- W4307531624 hasLocation W43075316241 @default.
- W4307531624 hasLocation W43075316242 @default.
- W4307531624 hasOpenAccess W4307531624 @default.
- W4307531624 hasPrimaryLocation W43075316241 @default.
- W4307531624 hasRelatedWork W1982615854 @default.
- W4307531624 hasRelatedWork W2014167307 @default.
- W4307531624 hasRelatedWork W2316250911 @default.
- W4307531624 hasRelatedWork W2754340464 @default.
- W4307531624 hasRelatedWork W2963009856 @default.
- W4307531624 hasRelatedWork W3131431199 @default.
- W4307531624 hasRelatedWork W4293824238 @default.
- W4307531624 hasRelatedWork W4304192548 @default.
- W4307531624 hasRelatedWork W4307531624 @default.
- W4307531624 hasRelatedWork W4380488735 @default.
- W4307531624 isParatext "false" @default.
- W4307531624 isRetracted "false" @default.
- W4307531624 workType "article" @default.