compneuroAISchool.py

import numpy as np
import scipy as sp
import math

#----1D input ------#

def oned(a):
  print("simulating a visual neuron's temporal response") 

  k=25;  #Temporal scale factor 
  n=5
  t = np.linspace(0,1,50)
  slow_t=np.power((k*t),n) * np.exp(-k*t)*(1/sp.math.factorial(n)-np.power((k*t),2)/sp.math.factorial(n+2))

  n = 3
  fast_t=np.power((k*t),n) * np.exp(-k*t)*(1/sp.math.factorial(n)-np.power((k*t),2)/sp.math.factorial(n+2))

  b = slow_t + fast_t # linear filter

  c=np.convolve(b,a) # convolve filter with stimulus
  d=c/np.max(c) # normalize

  e=100./(1.0 + np.exp(10*(0.5-d))) # static nonlinearity
  r=e; #for continuous output

  return(r)

#---- 3D input ------#

def threed(a):
  print("simulating a V1 simple cell's spatiotemporal response") 
  siz=a.shape
  x=siz[0]
  y=siz[1]
  z=siz[2]
  SIZE=x
  SF=0.15
  SIG=7
  OR=90*math.pi/180
  AR=3
  PH=0

  k=25;  #Temporal scale factor 
  n=5
  t = np.linspace(0,1,50)
  slow_t=np.power((k*t),n) * np.exp(-k*t)*(1/sp.math.factorial(n)-np.power((k*t),2)/sp.math.factorial(n+2))

  n = 3
  fast_t=np.power((k*t),n) * np.exp(-k*t)*(1/sp.math.factorial(n)-np.power((k*t),2)/sp.math.factorial(n+2))

  b = slow_t + fast_t # linear filter

  xx = np.linspace(1, SIZE, SIZE)
  xdata,ydata=np.meshgrid(xx,xx)
  temp1=(xdata-SIZE/2)*math.cos(OR)+(np.matrix.transpose(xdata)-SIZE/2)*np.sin(OR)
  temp2=(-xdata+SIZE/2)*math.sin(OR)+(np.matrix.transpose(xdata)-SIZE/2)*np.cos(OR)

  f1 = np.exp(-(np.multiply(temp1,temp1) + AR*AR*np.multiply(temp2,temp2)) / (2*SIG*SIG))
  f2 = np.cos(2*np.pi*SF*temp2+PH)
 
  rf_image = np.multiply(f1,f2)
  rf_image = rf_image - np.mean(rf_image)
  dp=[0]*(z)
  for ii in range (0,z-1,1):
	dp[ii]=np.sum(np.multiply(a[:,:,ii],rf_image))

  c=np.convolve(b,dp) # convolve filter with stimulus
  c=c/np.max(c) # normalize
  
  c[np.where(c<0)]=0

  r=c
  r[np.where(r>0.4)]=1 #spiking binary output
  r[np.where(r<=0.4)]=0
   


  return(r)