Projecten per jaar
Samenvatting
We introduce an Artificial Neural Network (ANN) quantization methodology for platforms without wide accumulation registers. This enables fixed-point model deployment on embedded compute platforms that are not specifically designed for large kernel computations (i.e. accumulator-constrained processors). We formulate the quantization problem as a function of accumulator size, and aim to maximize the model accuracy by maximizing bit width of input data and weights. To reduce the number of configurations to consider, only solutions that fully utilize the available accumulator bits are being tested. We demonstrate that 16 bit accumulators are able to obtain a classification accuracy within 1% of the floating-point baselines on the CIFAR-10 and ILSVRC2012 image classification benchmarks. Additionally, a near-optimal 2 × speedup is obtained on an ARM processor, by exploiting 16 bit accumulators for image classification on the All-CNN-C and AlexNet networks.
Originele taal-2 | Engels |
---|---|
Artikelnummer | 102872 |
Aantal pagina's | 11 |
Tijdschrift | Microprocessors and Microsystems |
Volume | 72 |
DOI's | |
Status | Gepubliceerd - 1 feb. 2020 |
Vingerafdruk
Duik in de onderzoeksthema's van 'Quantization of deep neural networks for accumulator-constrained processors'. Samen vormen ze een unieke vingerafdruk.Projecten
- 2 Afgelopen
-
Wearable Brainwave Processing Platform
Bergmans, J. W. M. (Project Manager), van der Hagen, D. (Project communicatie medewerker), Sánchez Martín, V. (Program Manager), Corporaal, H. (Projectmedewerker), Pineda de Gyvez, J. (Projectmedewerker) & Huisken, J. A. (Projectmedewerker)
1/09/16 → 30/11/21
Project: Onderzoek direct
-
Brainwave
Huisken, J. A. (Projectmedewerker), Jiao, H. (Project Manager), Singh, K. (Projectmedewerker), Sánchez Martín, V. (Project Manager), de Bruin, B. (Projectmedewerker), van der Hagen, D. (Project communicatie medewerker) & de Mol-Regels, M. (Project communicatie medewerker)
1/09/16 → 30/11/21
Project: Onderzoek direct