def outer_small(p1, fac_array, Ai, spi, posi, negi, likes_now):
N = fac_array.shape[1]
# For each sample from the posterior of r
for j in xrange(N):
# Compute actual PfPR[A[i]]
p2 = fac_array[:,j][Ai]
# Record likelihood
# k1=np.add.outer(np.log(p1),np.log(p2))
k1 = np.log(p1)*spi + np.dot(posi,np.log(p2))
# k2 = log(1.-np.outer(p1,p2))
k2 = cfh(p1,p2,negi)
# k1 = np.dot(k1,posi)
# k2 = np.dot(k2,negi)
likes_now[j,:] = k1 + k2
# Average log-likelihoods.
return np.apply_along_axis(logsum,0,likes_now) - log(N)#(mean(likes_now,axis=0))
def outer_small(p1, fac_array, Ai, spi, posi, negi, likes_now):
N = fac_array.shape[1]
# For each sample from the posterior of r
for j in xrange(N):
# Compute actual PfPR[A[i]]
p2 = fac_array[:, j][Ai]
# Record likelihood
# k1=np.add.outer(np.log(p1),np.log(p2))
k1 = np.log(p1) * spi + np.dot(posi, np.log(p2))
# k2 = log(1.-np.outer(p1,p2))
k2 = cfh(p1, p2, negi)
# k1 = np.dot(k1,posi)
# k2 = np.dot(k2,negi)
likes_now[j, :] = k1 + k2
# Average log-likelihoods.
return np.apply_along_axis(logsum, 0, likes_now) - log(
N) #(mean(likes_now,axis=0))
def this_fun(x, p2=p2, p3=p3,negi=negi, posi=posi, Ai=Ai):
p1 = np.log(invlogit(x))
return p1*spi + p3 + cfh(p1,p2,negi)
def this_fun(x, p2=p2, p3=p3, negi=negi, posi=posi, Ai=Ai):
p1 = np.log(invlogit(x))
return p1 * spi + p3 + cfh(p1, p2, negi)
