Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computing · HackerPapers discussions

Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computing — mentioned by p1esk

Thu, 16 Feb 2017 22:00:51 GMT

https://arxiv.org/abs/1603.08270 They have been designed, and are being used either for more efficient pattern matching, or to speed up brain simulations (again, using known neuronal models). You seem to expect something else from neuromorphic computing, why?

Paper and discussion history. A mention is not an endorsement.

Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computing — mentioned by deepnotderp

Sat, 11 Feb 2017 21:41:13 GMT

Yeah, this all sounds nice in practice, but look at the actual numbers: https://arxiv.org/abs/1603.08270 1) Nothing on imagenet 2) They already fall to 83% accuracy on CIFAR-10! Imagine how bad imagenet will be! If they string many chips together (exploding their power consumption, since here comes the Von Neumann Bottleneck of data movement), they get a paltry 89%..... Meanwhile even squeezenet achieves better results.

Paper and discussion history. A mention is not an endorsement.

Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computing — mentioned by emcq

Sat, 11 Feb 2017 21:27:32 GMT

The reality is that TrueNorth can use CNNs and be trained with backprop to achieve good performance with networks like ImageNet despite LeCun's comments [0, 1] LeCun's primary critique is that binary wont work; "to get good results on a task like ImageNet you need about 8 bit of precision on the neuron states". There was no evidence for his claim then, and now this is clearly false [2, 3]. LeCun's post is based more in pride than reason; he spends most of the time talking about NeuFlow which at one point was a competitor to True North for funding. In the end, NeuFlow never became a chip, but…

Paper and discussion history. A mention is not an endorsement.

Convolutional Networks for Fast, Energy-Efficient Neuromorphic Computing — mentioned by emcq

Sat, 11 Feb 2017 21:08:34 GMT

For training networks, designing en efficient chip is a scary proposition. However for inference tasks, low precision GEMM (as you said) goes a long way and better than what you often get. That's why chips like Movidius' Myriad are getting popular and are more similar to DSPs than neuromorphic designs. I agree that Intel's neuromorphic group doesn't get it, but other groups have taken neuromophic design principles that lead to efficient designs. For example, TrueNorth is very low precision, has great data locality, and though it was designed over 5 years ago can still use modern convolutional…

Paper and discussion history. A mention is not an endorsement.