Matches in SemOpenAlex for { <https://semopenalex.org/work/W4285274380> ?p ?o ?g. }
- W4285274380 endingPage "3254" @default.
- W4285274380 startingPage "3243" @default.
- W4285274380 abstract "The relentless growth of Artificial Intelligence (AI) workloads has fueled the drive towards non-Von Neuman architectures and custom computing hardware. Neuromorphic photonic engines aspire to synergize the low-power and high-bandwidth credentials of light-based deployments with novel architectures, towards surpassing the computing performance of their electronic counterparts. In this paper, we review recent progress in integrated photonic neuromorphic architectures and analyze the architectural and photonic hardware-based factors that limit their performance. Subsequently, we present our approach towards transforming silicon coherent neuromorphic layouts into high-speed and high-accuracy Deep Learning (DL) engines by combining robust architectures with hardware-aware DL training. Circuit robustness is ensured through a crossbar layout that circumvents insertion loss and fidelity constraints of state-of-the-art linear optical designs. Concurrently, we employ DL training models adapted to the underlying photonic hardware, incorporating noise- and bandwidth-limitations together with the supported activation function directly into Neural Network (NN) training. We validate experimentally the high-speed and high-accuracy advantages of hardware-aware DL models when combined with robust architectures through a SiPho prototype implementing a single column of a 4:4 photonic crossbar. This was utilized as the pen-ultimate hidden layer of a NN, revealing up to 5.93% accuracy improvement at 5GMAC/sec/axon when noise-aware training is enforced and allowing accuracies of 99.15% and 79.8% for the MNIST and CIFAR-10 classification tasks. Channel-aware training was then demonstrated by integrating the frequency response of the photonic hardware in NN training, with its experimental validation with the MNIST dataset revealing an accuracy increase of 12.93% at a record-high rate of 25GMAC/sec/axon." @default.
- W4285274380 created "2022-07-14" @default.
- W4285274380 creator A5021225470 @default.
- W4285274380 creator A5027421185 @default.
- W4285274380 creator A5037482385 @default.
- W4285274380 creator A5039953793 @default.
- W4285274380 creator A5041054091 @default.
- W4285274380 creator A5046173971 @default.
- W4285274380 creator A5046733238 @default.
- W4285274380 creator A5048648254 @default.
- W4285274380 creator A5061050264 @default.
- W4285274380 creator A5063280455 @default.
- W4285274380 creator A5067089800 @default.
- W4285274380 creator A5068387703 @default.
- W4285274380 date "2022-05-15" @default.
- W4285274380 modified "2023-10-10" @default.
- W4285274380 title "Neuromorphic Silicon Photonics and Hardware-Aware Deep Learning for High-Speed Inference" @default.
- W4285274380 cites W1992032065 @default.
- W4285274380 cites W2076476236 @default.
- W4285274380 cites W2117539524 @default.
- W4285274380 cites W2550362495 @default.
- W4285274380 cites W2752849906 @default.
- W4285274380 cites W2906829733 @default.
- W4285274380 cites W2919115771 @default.
- W4285274380 cites W2919727602 @default.
- W4285274380 cites W2923894441 @default.
- W4285274380 cites W2953409933 @default.
- W4285274380 cites W2957220981 @default.
- W4285274380 cites W2962730419 @default.
- W4285274380 cites W2963788845 @default.
- W4285274380 cites W2974263490 @default.
- W4285274380 cites W2977468072 @default.
- W4285274380 cites W2982130857 @default.
- W4285274380 cites W2982846042 @default.
- W4285274380 cites W2998986365 @default.
- W4285274380 cites W3006299272 @default.
- W4285274380 cites W3010046915 @default.
- W4285274380 cites W3010541957 @default.
- W4285274380 cites W3010964076 @default.
- W4285274380 cites W3026752472 @default.
- W4285274380 cites W3035921963 @default.
- W4285274380 cites W3049002444 @default.
- W4285274380 cites W3087893936 @default.
- W4285274380 cites W3103800690 @default.
- W4285274380 cites W3104406190 @default.
- W4285274380 cites W3117593917 @default.
- W4285274380 cites W3118265437 @default.
- W4285274380 cites W3120165331 @default.
- W4285274380 cites W3121141908 @default.
- W4285274380 cites W3128451613 @default.
- W4285274380 cites W3135765004 @default.
- W4285274380 cites W3138900639 @default.
- W4285274380 cites W3155628184 @default.
- W4285274380 cites W3191546639 @default.
- W4285274380 cites W3201727647 @default.
- W4285274380 cites W3203981982 @default.
- W4285274380 cites W3217333547 @default.
- W4285274380 cites W4226119313 @default.
- W4285274380 doi "https://doi.org/10.1109/jlt.2022.3171831" @default.
- W4285274380 hasPublicationYear "2022" @default.
- W4285274380 type Work @default.
- W4285274380 citedByCount "22" @default.
- W4285274380 countsByYear W42852743802022 @default.
- W4285274380 countsByYear W42852743802023 @default.
- W4285274380 crossrefType "journal-article" @default.
- W4285274380 hasAuthorship W4285274380A5021225470 @default.
- W4285274380 hasAuthorship W4285274380A5027421185 @default.
- W4285274380 hasAuthorship W4285274380A5037482385 @default.
- W4285274380 hasAuthorship W4285274380A5039953793 @default.
- W4285274380 hasAuthorship W4285274380A5041054091 @default.
- W4285274380 hasAuthorship W4285274380A5046173971 @default.
- W4285274380 hasAuthorship W4285274380A5046733238 @default.
- W4285274380 hasAuthorship W4285274380A5048648254 @default.
- W4285274380 hasAuthorship W4285274380A5061050264 @default.
- W4285274380 hasAuthorship W4285274380A5063280455 @default.
- W4285274380 hasAuthorship W4285274380A5067089800 @default.
- W4285274380 hasAuthorship W4285274380A5068387703 @default.
- W4285274380 hasConcept C104317684 @default.
- W4285274380 hasConcept C108583219 @default.
- W4285274380 hasConcept C113775141 @default.
- W4285274380 hasConcept C118524514 @default.
- W4285274380 hasConcept C120665830 @default.
- W4285274380 hasConcept C121332964 @default.
- W4285274380 hasConcept C127413603 @default.
- W4285274380 hasConcept C151927369 @default.
- W4285274380 hasConcept C154945302 @default.
- W4285274380 hasConcept C173608175 @default.
- W4285274380 hasConcept C185592680 @default.
- W4285274380 hasConcept C190502265 @default.
- W4285274380 hasConcept C20788544 @default.
- W4285274380 hasConcept C24326235 @default.
- W4285274380 hasConcept C2776257435 @default.
- W4285274380 hasConcept C29984679 @default.
- W4285274380 hasConcept C41008148 @default.
- W4285274380 hasConcept C50644808 @default.
- W4285274380 hasConcept C55493867 @default.
- W4285274380 hasConcept C63479239 @default.
- W4285274380 hasConcept C68339613 @default.