TY - UNPB
T1 - Neuromorphic Hardware In The Loop: Training a Deep Spiking Network on the BrainScaleS Wafer-Scale System
AU - Schmitt, Sebastian
AU - Klähn, Johann
AU - Bellec, Guillaume Emmanuel Fernand
AU - Grübl, Andreas
AU - Maurice, Güttler
AU - Hartl, Andreas
AU - Hartmann, Stephan
AU - Husmann, Dan
AU - Husmann, Kai
AU - Jeltsch, Sebastian
AU - Karasenko, Vitali
AU - Kleider, Mitja
AU - Koke, Christoph
AU - Kononov, Alexander
AU - Mauch, Christian
AU - Müller, Eric
AU - Müller, Paul
AU - Partzsch, Johannes
AU - Petrovici, Mihai
AU - Schiefer, Stefan
AU - Scholze, Stefan
AU - Thanasoulis, Vasilis
AU - Vogginger, Bernhard
AU - Legenstein, Robert
AU - Maass, Wolfgang
AU - Mayr, Christian
AU - Schüffny, Rene
AU - Schemmel, Johannes
AU - Meier, Karlheinz
PY - 2017/3/17
Y1 - 2017/3/17
N2 - Emulating spiking neural networks on analog neuromorphic hardware offers several advantages over simulating them on conventional computers, particularly in terms of speed and energy consumption. However, this usually comes at the cost of reduced control over the dynamics of the emulated networks. In this paper, we demonstrate how iterative training of a hardware-emulated network can compensate for anomalies induced by the analog substrate. We first convert a deep neural network trained in software to a spiking network on the BrainScaleS wafer-scale neuromorphic system, thereby enabling an acceleration factor of 10 000 compared to the biological time domain. This mapping is followed by the in-the-loop training, where in each training step, the network activity is first recorded in hardware and then used to compute the parameter updates in software via backpropagation. An essential finding is that the parameter updates do not have to be precise, but only need to approximately follow the correct gradient, which simplifies the computation of updates. Using this approach, after only several tens of iterations, the spiking network shows an accuracy close to the ideal software-emulated prototype. The presented techniques show that deep spiking networks emulated on analog neuromorphic devices can attain good computational performance despite the inherent variations of the analog substrate.
AB - Emulating spiking neural networks on analog neuromorphic hardware offers several advantages over simulating them on conventional computers, particularly in terms of speed and energy consumption. However, this usually comes at the cost of reduced control over the dynamics of the emulated networks. In this paper, we demonstrate how iterative training of a hardware-emulated network can compensate for anomalies induced by the analog substrate. We first convert a deep neural network trained in software to a spiking network on the BrainScaleS wafer-scale neuromorphic system, thereby enabling an acceleration factor of 10 000 compared to the biological time domain. This mapping is followed by the in-the-loop training, where in each training step, the network activity is first recorded in hardware and then used to compute the parameter updates in software via backpropagation. An essential finding is that the parameter updates do not have to be precise, but only need to approximately follow the correct gradient, which simplifies the computation of updates. Using this approach, after only several tens of iterations, the spiking network shows an accuracy close to the ideal software-emulated prototype. The presented techniques show that deep spiking networks emulated on analog neuromorphic devices can attain good computational performance despite the inherent variations of the analog substrate.
M3 - Preprint
VL - arXiv:1703.01909
T3 - arXiv.org e-Print archive
BT - Neuromorphic Hardware In The Loop: Training a Deep Spiking Network on the BrainScaleS Wafer-Scale System
ER -