Why can't I use this code for the SHD dataset as accurate as in the paper?

idiap / sparch

PyTorch based toolkit for developing spiking neural networks (SNNs) by training and testing them on speech command recognition tasks

20 stars 4 forks source link

Why can't I use this code for the SHD dataset as accurate as in the paper? #5

Open dgxdn opened 1 month ago

dgxdn commented 1 month ago

LIF、RLIF、adLIF、RadLIF

dzcgood commented 1 month ago

LIF、RLIF、adLIF、RadLIF

I recently noticed this paper and found that the accuracy is very high, but I haven't tried to reproduce it yet, could you please tell me what is the accuracy of your reproduction?

dgxdn commented 1 month ago

I want to use the code to reproduce its results on SHD, and both Radlif and adlif are less than 92% accurate