AUTHOR=Stromatias Evangelos , Soto Miguel , Serrano-Gotarredona Teresa , Linares-Barranco Bernabé 

TITLE=An Event-Driven Classifier for Spiking Neural Networks Fed with Synthetic or Dynamic Vision Sensor Data

JOURNAL=Frontiers in Neuroscience

VOLUME=Volume 11 - 2017

YEAR=2017

URL=https://www.frontiersin.org/journals/neuroscience/articles/10.3389/fnins.2017.00350

DOI=10.3389/fnins.2017.00350

ISSN=1662-453X

ABSTRACT=This paper introduces a novel methodology for training an event-based classifier with synthetic and raw dynamic vision sensor (DVS) data. The proposed supervised method takes advantage of the spiking activity to build histograms and train the classifier in the frame domain using the stochastic gradient descend algorithm. In addition, this approach can cope with neuron leakages, a desirable feature for real-world applications, since it captures the dynamics of the spikes. We tested our method on the MNIST data set using different encodings and DVS-based data sets such as N-MNIST, MNIST-DVS, and Fast-Poker-DVS using the same network topology and feature maps. We demonstrate the effectiveness of our approach by achieving the highest classification accuracy reported on the N-MNIST to date with a spiking convolutional network 97.77%, as well as, 100% on the Fast-Poker-DVS data set. Moreover, by using the proposed method we were able to retrain the output layer  of a spiking neural network and increase its performance by 2% suggesting that our classifier can be used as the output layer in works where features are extracted using unsupervised spike-based learning methods. Lastly, this work also presents a comparison between different data sets in terms of total activity and network latency.