Fix spike/membrane potential independence

[leoauri]

Here is a way that NoisyLeaky can work, resetting the membrane potential if and only if the neuron spikes.

Demo:

n = NoisyLeaky(beta=0.5, output=True, reset_mechanism="zero")

steps = 20
outs = []
mem = torch.zeros(1)
for t in range(steps):
    out, mem = n(torch.Tensor([[1.0]]), mem)
    outs.append((out, mem))

outs

lif2 = NoisyLeaky(beta=0.6)

# Initialize inputs and outputs
cur_in = torch.cat((torch.zeros(10, 1), torch.ones(190, 1)*0.2), 0)
mem = torch.zeros(1)
spk_out = torch.zeros(1) 
mem_rec = [mem]
spk_rec = [spk_out]

for ci in cur_in:
  spk_out, mem = lif2(ci, mem)
  mem_rec.append(mem)
  spk_rec.append(spk_out)

# convert lists to tensors
mem_rec = torch.stack(mem_rec)
spk_rec = torch.stack(spk_rec)

plot_cur_mem_spk(cur_in, mem_rec, spk_rec, thr_line=1, vline=109, ylim_max2=1.3)

That seems to make more sense to me.

I decided to change the mem output to reflect the value after spike/membrane reset in that step. Otherwise, I would have had to either make the class stateful or change the interface. Also, this just makes more sense to my brain. How does that play with the rest of the SNN world?

Cheers,
Leo

[leoauri]

Fixes #366