Papers of the day   All papers

SLIDE : IN DEFENSE OF SMART ALGORITHMS OVER HARDWARE ACCELERATION FOR LARGE-SCALE DEEP LEARNING SYSTEMS

Comments

Brandon Rohrer: Brilliant work on modifying neural networks to run faster on CPU than on GPU. The trick is sparsity, multithreading, and carefully managing overhead. GPUs are great, but you don't need them to start playing with neural networks. https://arxiv.org/abs/1903.03129

2 replies, 329 likes


Hacker News: Slide: 3.5x Faster Deep Learning on CPU Then on GPU https://arxiv.org/abs/1903.03129

1 replies, 63 likes


Hacker News: Smart Algorithms beat Hardware Acceleration for Large-Scale Deep Learning https://arxiv.org/abs/1903.03129

0 replies, 31 likes


Hacker News: Slide: 3.5x Faster Deep Learning on CPU Then on GPU https://arxiv.org/abs/1903.03129

0 replies, 13 likes


Anshumali: Smart Algorithm over CPU -VS- Tensorflow(@TensorFlo) over #V100 GPU? Check it out: Codes: https://github.com/RUSH-LAB/HashingDeepLearning Paper: https://arxiv.org/abs/1903.03129 Deep learning packages need hardware because maybe we don't write ML codes from scratch. #HPC #DeepLearning #ML #AI #BigData #GPU

0 replies, 11 likes


HPC Guru: SLIDE: Smart randomized algorithms on #CPU can outperform the best available #GPU with an optimized implementation. “#DeepLearning packages need hardware because maybe we don't write ML codes from scratch” @RiceUniversity https://arxiv.org/pdf/1903.03129.pdf #AI #HPC via @Anshumali_ https://t.co/pyiFp8JSxv

1 replies, 10 likes


Karen: Oh these people are going to get it...I'm still on the Intro and already going full Reviewer 3... https://arxiv.org/abs/1903.03129

1 replies, 7 likes


arxiv: SLIDE : In Defense of Smart Algorithms over Hardware Acceleration for Large-Scale Deep Le... http://arxiv.org/abs/1903.03129 https://t.co/YOp8AerG3t

0 replies, 5 likes


Hacker News 100: Slide: 3.5x Faster Deep Learning on CPU Then on GPU https://arxiv.org/abs/1903.03129 (http://bit.ly/2vLG9hT)

0 replies, 4 likes


Nige Willson: Deep learning rethink overcomes major obstacle in AI industry http://bit.ly/2PFGo4Z #ai #ArtificialIntelligence #MachineLearning #DeepLearning https://t.co/XHVZBitIip

0 replies, 3 likes


Cosmin Stamate: SLIDE: smart algorithms can make large scale deep learning systems run faster on CPUs than GPUs. https://arxiv.org/abs/1903.03129 Fresh off the Intel press! https://t.co/KMcodrScvh

0 replies, 3 likes


Stephan Schmidt 🚀: CPU Deep Learning 3.5x faster than GPU https://arxiv.org/abs/1903.03129

0 replies, 1 likes


M157q News RSS: Smart Algorithms beat Hardware Acceleration for Large-Scale Deep Learning https://arxiv.org/abs/1903.03129 Article URL: https://arxiv.org/abs/1903.03129 Comments URL: https://news.ycombinator.com/item?id=22502131 Points: 114 # Comments: 80

0 replies, 1 likes


imuno@後衛ミニオン: https://arxiv.org/pdf/1903.03129.pdf https://github.com/keroro824/HashingDeepLearning

0 replies, 1 likes


HPC Guru: @evanburness Good point, Evan. I am a big fan of improving algorithms for better performance. @NicoleHemsoth interviewed one of the authors on Next Platform Radio yesterday. However, @KpDooty has raised some concerns about this paper. https://twitter.com/KpDooty/status/1234850011797651456?s=20 #HPC #AI #DNNs

1 replies, 0 likes


Content

Found on Mar 07 2020 at https://arxiv.org/pdf/1903.03129.pdf

PDF content of a computer science paper: SLIDE : IN DEFENSE OF SMART ALGORITHMS OVER HARDWARE ACCELERATION FOR LARGE-SCALE DEEP LEARNING SYSTEMS