Matches in SemOpenAlex for { <https://semopenalex.org/work/W2948252829> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W2948252829 abstract "Deep Convolutional Neural Networks (CNNs) now match human accuracy in many image prediction tasks, resulting in a growing adoption in e-commerce, radiology, and other domains. Naturally, explaining CNN predictions is a key concern for many users. Since the internal workings of CNNs are unintuitive for most users, occlusion-based explanations (OBE) are popular for understanding which parts of an image matter most for a prediction. One occludes a region of the image using a patch and moves it around to produce a heat map of changes to the prediction probability. Alas, this approach is computationally expensive due to the large number of re-inference requests produced, which wastes time and raises resource costs. We tackle this issue by casting the OBE task as a new instance of the classical incremental view maintenance problem. We create a novel and comprehensive algebraic framework for incremental CNN inference combining materialized views with multi-query optimization to reduce computational costs. We then present two novel approximate inference optimizations that exploit the semantics of CNNs and the OBE task to further reduce runtimes. We prototype our ideas in Python to create a tool we call Krypton that supports both CPUs and GPUs. Experiments with real data and CNNs show that Krypton reduces runtimes by up to 5X (resp. 35X) to produce exact (resp. high-quality approximate) results without raising resource requirements." @default.
- W2948252829 created "2019-06-14" @default.
- W2948252829 creator A5059167302 @default.
- W2948252829 creator A5069604107 @default.
- W2948252829 creator A5075059497 @default.
- W2948252829 date "2019-06-25" @default.
- W2948252829 modified "2023-10-12" @default.
- W2948252829 title "Incremental and Approximate Inference for Faster Occlusion-based Deep CNN Explanations" @default.
- W2948252829 cites W1849277567 @default.
- W2948252829 cites W1975733288 @default.
- W2948252829 cites W2026102093 @default.
- W2948252829 cites W2058978608 @default.
- W2948252829 cites W2086140762 @default.
- W2948252829 cites W2108598243 @default.
- W2948252829 cites W2117539524 @default.
- W2948252829 cites W2133665775 @default.
- W2948252829 cites W2282821441 @default.
- W2948252829 cites W2606822607 @default.
- W2948252829 cites W2613621736 @default.
- W2948252829 cites W2752236330 @default.
- W2948252829 cites W2784477424 @default.
- W2948252829 cites W2788633781 @default.
- W2948252829 cites W2962858109 @default.
- W2948252829 cites W2963363373 @default.
- W2948252829 cites W2963387679 @default.
- W2948252829 cites W2963893493 @default.
- W2948252829 cites W3098547670 @default.
- W2948252829 doi "https://doi.org/10.1145/3299869.3319874" @default.
- W2948252829 hasPublicationYear "2019" @default.
- W2948252829 type Work @default.
- W2948252829 sameAs 2948252829 @default.
- W2948252829 citedByCount "26" @default.
- W2948252829 countsByYear W29482528292019 @default.
- W2948252829 countsByYear W29482528292020 @default.
- W2948252829 countsByYear W29482528292021 @default.
- W2948252829 countsByYear W29482528292022 @default.
- W2948252829 countsByYear W29482528292023 @default.
- W2948252829 crossrefType "proceedings-article" @default.
- W2948252829 hasAuthorship W2948252829A5059167302 @default.
- W2948252829 hasAuthorship W2948252829A5069604107 @default.
- W2948252829 hasAuthorship W2948252829A5075059497 @default.
- W2948252829 hasBestOaLocation W29482528291 @default.
- W2948252829 hasConcept C108583219 @default.
- W2948252829 hasConcept C119857082 @default.
- W2948252829 hasConcept C154945302 @default.
- W2948252829 hasConcept C162324750 @default.
- W2948252829 hasConcept C187736073 @default.
- W2948252829 hasConcept C199360897 @default.
- W2948252829 hasConcept C2776214188 @default.
- W2948252829 hasConcept C2777472644 @default.
- W2948252829 hasConcept C2780451532 @default.
- W2948252829 hasConcept C41008148 @default.
- W2948252829 hasConcept C519991488 @default.
- W2948252829 hasConcept C80444323 @default.
- W2948252829 hasConcept C81363708 @default.
- W2948252829 hasConceptScore W2948252829C108583219 @default.
- W2948252829 hasConceptScore W2948252829C119857082 @default.
- W2948252829 hasConceptScore W2948252829C154945302 @default.
- W2948252829 hasConceptScore W2948252829C162324750 @default.
- W2948252829 hasConceptScore W2948252829C187736073 @default.
- W2948252829 hasConceptScore W2948252829C199360897 @default.
- W2948252829 hasConceptScore W2948252829C2776214188 @default.
- W2948252829 hasConceptScore W2948252829C2777472644 @default.
- W2948252829 hasConceptScore W2948252829C2780451532 @default.
- W2948252829 hasConceptScore W2948252829C41008148 @default.
- W2948252829 hasConceptScore W2948252829C519991488 @default.
- W2948252829 hasConceptScore W2948252829C80444323 @default.
- W2948252829 hasConceptScore W2948252829C81363708 @default.
- W2948252829 hasLocation W29482528291 @default.
- W2948252829 hasOpenAccess W2948252829 @default.
- W2948252829 hasPrimaryLocation W29482528291 @default.
- W2948252829 hasRelatedWork W2032775418 @default.
- W2948252829 hasRelatedWork W2058978608 @default.
- W2948252829 hasRelatedWork W2099102906 @default.
- W2948252829 hasRelatedWork W2357449897 @default.
- W2948252829 hasRelatedWork W2432911982 @default.
- W2948252829 hasRelatedWork W2548122763 @default.
- W2948252829 hasRelatedWork W2743948853 @default.
- W2948252829 hasRelatedWork W2769041395 @default.
- W2948252829 hasRelatedWork W2791094827 @default.
- W2948252829 hasRelatedWork W2889073286 @default.
- W2948252829 hasRelatedWork W2891867110 @default.
- W2948252829 hasRelatedWork W2946595616 @default.
- W2948252829 hasRelatedWork W2952926545 @default.
- W2948252829 hasRelatedWork W2963576971 @default.
- W2948252829 hasRelatedWork W2963815651 @default.
- W2948252829 hasRelatedWork W2977366063 @default.
- W2948252829 hasRelatedWork W2987732534 @default.
- W2948252829 hasRelatedWork W3081844508 @default.
- W2948252829 hasRelatedWork W3106020963 @default.
- W2948252829 hasRelatedWork W3162616539 @default.
- W2948252829 isParatext "false" @default.
- W2948252829 isRetracted "false" @default.
- W2948252829 magId "2948252829" @default.
- W2948252829 workType "article" @default.