Export Reviews, Discussions, Author Feedback and Meta-Reviews

The authors present an interesting paper about a backpropagation training method using spike probabilities and takes into account the hardware constraints of a particular platform which has spiking neurons and discrete synapses. This is a topic of current interest in mapping deep networks to multi-neuron hardware platforms. Quality: The proposed training method is useful especially in consideration of new multi-neuron hardware platforms with constraints. Clarity: The paper is easy to read. The claims made at the end of Section 1 can be reworded because as it stands, if not read carefully, suggests that the paper proposes for the first time a training methodology that employs spiking neurons, synapses with reduced precision.