Matches in SemOpenAlex for { <https://semopenalex.org/work/W3177005201> ?p ?o ?g. }
- W3177005201 abstract "While multi-modal deep learning is useful in distributed sensing tasks like human tracking, activity recognition, and audio and video analysis, deploying state-of-the-art multi-modal models in a wirelessly networked sensor system poses unique challenges. The data sizes for different modalities can be highly asymmetric (e.g., video vs. audio), and these differences can lead to significant delays between streams in the presence of wireless dynamics. Therefore, a slow stream can significantly slow down a multi-modal inference system in the cloud, leading to either increased latency (when blocked by the slow stream) or degradation in inference accuracy (if inference proceeds without waiting). In this paper, we introduce speculative inference on multi-modal data streams to adapt to these asymmetries across modalities. Rather than blocking inference until all sensor streams have arrived and been temporally aligned, we impute any missing, corrupt, or partially-available sensor data, then generate a speculative inference using the learned models and imputed data. A rollback module looks at the class output of speculative inference and determines whether the class is sufficiently robust to incomplete data to accept the result; if not, we roll back the inference and update the model's output. We implement the system in three multi-modal application scenarios using public datasets. The experimental results show that our system achieves 7 -- 128× latency speedup with the same accuracy as six state-of-the-art methods." @default.
- W3177005201 created "2021-07-05" @default.
- W3177005201 creator A5023901511 @default.
- W3177005201 creator A5069858475 @default.
- W3177005201 creator A5074602782 @default.
- W3177005201 creator A5086353365 @default.
- W3177005201 date "2021-06-24" @default.
- W3177005201 modified "2023-10-12" @default.
- W3177005201 title "Low-latency speculative inference on distributed multi-modal data streams" @default.
- W3177005201 cites W1526392145 @default.
- W3177005201 cites W1840179877 @default.
- W3177005201 cites W2015143272 @default.
- W3177005201 cites W2025549137 @default.
- W3177005201 cites W2054242451 @default.
- W3177005201 cites W2057907879 @default.
- W3177005201 cites W2062118960 @default.
- W3177005201 cites W2080501585 @default.
- W3177005201 cites W2086402015 @default.
- W3177005201 cites W2100495367 @default.
- W3177005201 cites W2115438344 @default.
- W3177005201 cites W2116012015 @default.
- W3177005201 cites W2117229510 @default.
- W3177005201 cites W2121045468 @default.
- W3177005201 cites W2132585078 @default.
- W3177005201 cites W2144564081 @default.
- W3177005201 cites W2148298736 @default.
- W3177005201 cites W2165253089 @default.
- W3177005201 cites W2194775991 @default.
- W3177005201 cites W2331128040 @default.
- W3177005201 cites W2335722297 @default.
- W3177005201 cites W2434537049 @default.
- W3177005201 cites W2468875367 @default.
- W3177005201 cites W2518157807 @default.
- W3177005201 cites W2530120573 @default.
- W3177005201 cites W2566832195 @default.
- W3177005201 cites W2607541615 @default.
- W3177005201 cites W2621682591 @default.
- W3177005201 cites W2623638694 @default.
- W3177005201 cites W2734326504 @default.
- W3177005201 cites W2744628735 @default.
- W3177005201 cites W2777460464 @default.
- W3177005201 cites W2786070938 @default.
- W3177005201 cites W2804787784 @default.
- W3177005201 cites W2887117815 @default.
- W3177005201 cites W2951611190 @default.
- W3177005201 cites W2963073614 @default.
- W3177005201 cites W2963311060 @default.
- W3177005201 cites W2963426391 @default.
- W3177005201 cites W2963654155 @default.
- W3177005201 cites W2963800363 @default.
- W3177005201 cites W2965204941 @default.
- W3177005201 cites W2969432809 @default.
- W3177005201 cites W3009217644 @default.
- W3177005201 cites W3018353706 @default.
- W3177005201 cites W3035574168 @default.
- W3177005201 cites W3035600682 @default.
- W3177005201 cites W3092134327 @default.
- W3177005201 cites W3103272053 @default.
- W3177005201 cites W3135892362 @default.
- W3177005201 cites W3151203111 @default.
- W3177005201 cites W3151769848 @default.
- W3177005201 cites W4210656804 @default.
- W3177005201 cites W4232009874 @default.
- W3177005201 cites W753847829 @default.
- W3177005201 doi "https://doi.org/10.1145/3458864.3467884" @default.
- W3177005201 hasPublicationYear "2021" @default.
- W3177005201 type Work @default.
- W3177005201 sameAs 3177005201 @default.
- W3177005201 citedByCount "7" @default.
- W3177005201 countsByYear W31770052012021 @default.
- W3177005201 countsByYear W31770052012022 @default.
- W3177005201 countsByYear W31770052012023 @default.
- W3177005201 crossrefType "proceedings-article" @default.
- W3177005201 hasAuthorship W3177005201A5023901511 @default.
- W3177005201 hasAuthorship W3177005201A5069858475 @default.
- W3177005201 hasAuthorship W3177005201A5074602782 @default.
- W3177005201 hasAuthorship W3177005201A5086353365 @default.
- W3177005201 hasConcept C104317684 @default.
- W3177005201 hasConcept C111919701 @default.
- W3177005201 hasConcept C124101348 @default.
- W3177005201 hasConcept C154945302 @default.
- W3177005201 hasConcept C173608175 @default.
- W3177005201 hasConcept C185592680 @default.
- W3177005201 hasConcept C188027245 @default.
- W3177005201 hasConcept C24590314 @default.
- W3177005201 hasConcept C2776214188 @default.
- W3177005201 hasConcept C2778484313 @default.
- W3177005201 hasConcept C31258907 @default.
- W3177005201 hasConcept C41008148 @default.
- W3177005201 hasConcept C55493867 @default.
- W3177005201 hasConcept C63479239 @default.
- W3177005201 hasConcept C68339613 @default.
- W3177005201 hasConcept C71139939 @default.
- W3177005201 hasConcept C76155785 @default.
- W3177005201 hasConcept C79403827 @default.
- W3177005201 hasConcept C79974875 @default.
- W3177005201 hasConcept C82876162 @default.
- W3177005201 hasConcept C89198739 @default.
- W3177005201 hasConceptScore W3177005201C104317684 @default.
- W3177005201 hasConceptScore W3177005201C111919701 @default.