Zobrazeno 1 - 10
of 97
pro vyhledávání: '"Jorge Albericio"'
Structured channel pruning has been shown to significantly accelerate inference time for convolution neural networks (CNNs) on modern hardware, with a relatively minor loss of network accuracy. Recent works permanently zero these channels during trai
Externí odkaz:
http://arxiv.org/abs/2211.02206
Autor:
Mishra, Asit, Latorre, Jorge Albericio, Pool, Jeff, Stosic, Darko, Stosic, Dusan, Venkatesh, Ganesh, Yu, Chong, Micikevicius, Paulius
As neural network model sizes have dramatically increased, so has the interest in various techniques to reduce their parameter counts and accelerate their execution. An active area of research in this field is sparsity - encouraging zero values in pa
Externí odkaz:
http://arxiv.org/abs/2104.08378
Publikováno v:
IEEE Transactions on Computers. 71:3072-3073
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::ff4b17a97d5b4329aeba99039a6eb414
https://doi.org/10.1109/tc.2022.3216428
https://doi.org/10.1109/tc.2022.3216428
Publikováno v:
Lecture Notes in Computer Science ISBN: 9783031200823
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::6ef0c55332121b79056fd5734574f49d
https://doi.org/10.1007/978-3-031-20083-0_38
https://doi.org/10.1007/978-3-031-20083-0_38
Autor:
Kevin Siu, Sayeh Sharify, Mostafa Mahmoud, Dylan Malone Stuart, Alberto Delmas Lascorz, Andreas Moshovos, Jorge Albericio, Isak Edo Vivancos, Patrick Judd, Zissis Poulos, Milos Nikolic
Publikováno v:
IEEE Micro. 39:26-35
We review two inference accelerators that exploit value properties in deep neural networks: 1) Diffy that targets spatially correlated activations in computational imaging DNNs, and 2) Tactical that targets sparse neural networks using a low-overhead
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::33d8d8a6a6228b5e48b284800ca62212
https://doi.org/10.1109/mm.2019.2930596
https://doi.org/10.1109/mm.2019.2930596
Autor:
Tor M. Aamodt, Alberto Delmas Lascorz, Tayler Hetherington, Patrick Judd, Zissis Poulos, Sayeh Sharify, Jorge Albericio, Andreas Moshovos, Natalie Enright Jerger
Publikováno v:
Computer. 51:18-30
To deliver the hardware computation power advances needed to support deep learning innovations, identifying deep learning properties that designers could potentially exploit is invaluable. This article articulates our strategy and overviews several v
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::3df1730612bd74a2a7fa4bae93412c09
https://doi.org/10.1109/mc.2018.2381114
https://doi.org/10.1109/mc.2018.2381114
Autor:
Jorge Albericio, Tayler Hetherington, Patrick Judd, Raquel Urtasun, Natalie Enright Jerger, Andreas Moshovos, Tor M. Aamodt
Publikováno v:
Parallel Computing. 73:40-51
This work investigates how using reduced precision data in Deep Neural Networks (DNNs) affects network accuracy during classification. We observe that the tolerance of DNNs to reduced precision data not only varies across networks, but also within ne
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::48b83a99335e594249ca3f80ac30f50e
https://doi.org/10.1016/j.parco.2017.05.003
https://doi.org/10.1016/j.parco.2017.05.003
Autor:
Patrick Judd, Tor M. Aamodt, Andreas Moshovos, Sayeh Sharify, Tayler Hetherington, Alberto Delmas Lascorz, Jorge Albericio, Natalie Enright Jerger
Publikováno v:
IEEE Micro. 38:41-55
This article summarizes our recent work on value-based hardware accelerators for image classification using Deep Convolutional Neural Networks (CNNs). The presented designs exploit runtime value properties that are difficult or impossible to discern
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::35fb49990f196cc846254810d57cb3e5
https://doi.org/10.1109/mm.2018.112130309
https://doi.org/10.1109/mm.2018.112130309
Publikováno v:
IEEE Computer Architecture Letters. 16:80-83
The numerical representation precision required by the computations performed by Deep Neural Networks (DNNs) varies across networks and between layers of a same network. This observation motivates a precision-based approach to acceleration which take
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::5e670bba5d06b6fcfaa53f41710018ec
https://doi.org/10.1109/lca.2016.2597140
https://doi.org/10.1109/lca.2016.2597140
Identifying and Exploiting Ineffectual Computations to Enable Hardware Acceleration of Deep Learning
Autor:
Kevin Siu, Sayeh Sharify, Tayler Hetherington, Mostafa Mahmoud, Alberto Delmas, Tor M. Aamodt, Natalie Enright Jerger, Zissis Poulos, Andreas Moshovos, Dylan Malone Stuart, Milos Nikolic, Jorge Albericio, Patrick Judd
Publikováno v:
NEWCAS
This article summarizes somde of our work on hardware accelerators for inference with Deep Learning Neural Networks (DNNs). Early success in hardware acceleration for DNNs exploited the computation structure and the significant reuse in their access
Externí odkaz:
https://explore.openaire.eu/search/publication?articleId=doi_________::61505651785664f5813f1872ebbd807b
https://doi.org/10.1109/newcas.2018.8585656
https://doi.org/10.1109/newcas.2018.8585656