def test_compute_kernel():
kernel_pnts = 100
new_kernel = np.zeros(kernel_pnts)
new_kernel[0] = 1
f_block = ns.filter_block(1, new_kernel, 1)
trials = 1000
contrast = 10
mean = 127
length = trials * kernel_pnts * ns.sim_delta_t
ker_length = kernel_pnts * ns.sim_delta_t
stim = ns.fake_noise('gaussian', contrast, length=length, mean=mean)
# simulate linear response
lp = f_block.temporal_filter(stim)
stim = stim[kernel_pnts-1:] # now stim and lp_center are aligned
# Recompute kernel from the stim and the lp
new_kernel = ns.compute_kernel(stim, lp, kernel_pnts,10)
plt.close('kernel_test')
fix, ax = plt.subplots(num='kernel_test')
ax.plot(f_block.kernel)
ax.plot(new_kernel)
return new_kernel, f_block.kernel, stim
def test_temporal_filter():
# make a kernel that is a delta function at t=0
new_kernel = np.zeros(100)
new_kernel[0] = 1
kernel = ns.filter_block(1, new_kernel, 1)
# make stim
stim = ns.fake_noise('gaussian', 10)
lp = kernel.temporal_filter(stim)
# First len(new_kernel)-1 points of stim were not properly filtered and are not included in lp
np.testing.assert_almost_equal(lp, stim[len(new_kernel)-1:])
