TRSasasusu/SpikingMoYF

Consistent with The Target Paper

Closed this issue · 3 comments

Deleted user commented

The way input spikes are generated is inconsistent with the description in the paper Training Deep Spiking Neural Networks Using Backpropagation.

TRSasasusu commented

Thank you for making an issue, but now I think this kind of SNN is not essential to real neuroscience, so I have no plan to fix it.

(However, I would like to know what is inconsistent with the description about input spikes generation. Could you tell me the detail?)

Deleted user commented

I totally agree that neuron parameters used in the paper are not realistic.

The author creates 784 Poisson spiking neurons; each of which corresponds to an individual pixel of MNIST sample. Pixel intensity (0-255) is scaled and set to be the firing rate of the Poisson spiking neurons. Thus, at any time instance, chance of firing depends on scaled value of pixel intensity.

TRSasasusu commented

Thank you! The link of this issue has just been placed on the top of README.