Abstract
On-chip edge intelligence has necessitated the exploration of algorithmic techniques to reduce the compute requirements of current machine learning frameworks. This work aims to bridge the recent algorithmic progress in training Binary Neural Networks and Spiking Neural Networks—both of which are driven by the same motivation and yet synergies between the two have not been fully explored. We show that training Spiking Neural Networks in the extreme quantization regime results in near full precision accuracies on large-scale datasets like CIFAR-100 and ImageNet. An important implication of this work is that Binary Spiking Neural Networks can be enabled by “In-Memory” hardware accelerators catered for Binary Neural Networks without suffering any accuracy degradation due to binarization. We utilize standard training techniques for non-spiking networks to generate our spiking networks by conversion process and also perform an extensive empirical analysis and explore simple design-time and run-time optimization techniques for reducing inference latency of spiking networks (both for binary and full-precision models) by an order of magnitude over prior work. Our implementation source code and trained models are available at https://github.com/NeuroCompLab-psu/SNN-Conversion.
Highlights
The explosive growth of edge devices such as mobile phones, wearables, smart sensors and robotic devices in the current Internet of Things (IoT) era has driven the research for the quest of machine learning platforms that are accurate but are optimal from storage and compute requirements
Since our Spiking Neural Networks” (SNNs) training optimizations are valid for full-precision networks, we report accuracies for full-precision models along with their binary counterparts in order to compare against prior art
We explored an “Early Exit” inference method for SNNs wherein we consider the SNN inference to be completed when the maximum membrane potential of the neurons in the final layer is above a given threshold3
Summary
The explosive growth of edge devices such as mobile phones, wearables, smart sensors and robotic devices in the current Internet of Things (IoT) era has driven the research for the quest of machine learning platforms that are accurate but are optimal from storage and compute requirements. On-device edge intelligence has become increasingly crucial with the advent of a plethora of applications that require real-time information processing with limited connectivity to cloud servers. Privacy concerns for data sharing with remote servers have fueled the need for on-chip intelligence in resourced constrained, battery-life limited edge devices. To address these challenges, a wide variety of works in the deep learning community have explored mechanisms for model compression like pruning (Han et al, 2015; Alvarez and Salzmann, 2017), efficient network architectures (Iandola et al, 2016), reduced precision/quantized networks (Hubara et al, 2017), among others. The drastic reductions in computation time result from the fact that costly Multiply-Accumulate operations required
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
