How to writte a raster plot?

user_14452992 0
Viewed 138

I'm using pytorch on google colab. I've got a tensor matrix below, this is the example, and actually the matrix size is about 50 neurons and 30,000~50,000 time.

a= torch.tensor([[0., 0., 0., 0., 0.],
                 [0., 0., 0., 0., 1.],
                 [0., 1., 0., 1., 0.]])

each of values are,

a= torch.tensor([[Neuron1(t=1), N2(t=1), N3(t=1), N4(t=1), N5(t=1)],
                 [N1(t=2), N2(t=2), N3(t=2), N4(t=2), N5(t=2)],
                 [N1(t=3), N2(t=3), N3(t=3), N4(t=3), N5(t=3)]])

and 1 means that neuron fire, 0 means not fire.
So Neuron5(t=2), Neuron2(t=3) and Neuron4(t=3) are firing.
Then, I want to make a raster plot or scatter plot like below using this matrix,
The dots show the firing neuron.

neuron number
1|
2|          *
3|
4|          *
5|__ *_____time
    1  2  3

What would be the best python code to do this? I have no idea now. Thank you for reading.

pytorch google-colaboratory python
3 Answers

You can do it easily as follows:

import matplotlib.pyplot as plt
a= torch.tensor([[0., 0., 0., 0., 0.],
                 [0., 0., 0., 0., 1.],
                 [0., 1., 0., 1., 0.]],device='cuda')
plt.scatter(*torch.where(a.cpu()))
user_47577150

High resolution version, derived from Mr Gil's code.

a= ~~~a huge torch.tensor (50 neurons and 30,000 time)~~~

fig = plt.figure(facecolor="w", figsize=(300,5))
ax = fig.add_subplot(111)
ax.scatter(*torch.where(a.cpu()),
                      s=0.05[![enter image description here][1]][1],
                      c="black",
                      linewidths="0")
plt.savefig("rascatter.png",format="png", dpi=120)
user_144529920

There are two answers here. Thank you very much everyone.

One is Mr Gil Pinsky and the other is Mr tgrandje. Mr Gil's code is on above. Thank you.

And here is the code I wrote based on Mr tgrandje's advice. 

a=input_list
b = a.to('cpu').detach().numpy().copy()
c=b.T
raster = plt.imshow(c,cmap="Greys",aspect=1)
# plt.show
plt.savefig("raster.png", format="png", dpi=900)
user_144529920
Related