def test_scale_design_mtx():
actual_design = np.ones((len(convolved), 2))
actual_design[:, 1] = convolved
exp_design , exp_B_4d = glm(data, convolved)
f1=scale_design_mtx(exp_design)[-1]
r1=np.array([ 1. , 0.16938989])
assert_almost_equal(f1,r1)
def test_scale_design_mtx():
actual_design = np.ones((len(convolved), 2))
actual_design[:, 1] = convolved
exp_design, exp_B_4d = glm(data, convolved)
f1 = scale_design_mtx(exp_design)[-1]
r1 = np.array([1., 0.16938989])
assert_almost_equal(f1, r1)
def test_glm1():
actual_design = np.ones((len(convolved), 2))
actual_design[:, 1] = convolved
data_2d = np.reshape(data, (-1, data.shape[-1]))
actual_B = npl.pinv(actual_design).dot(data_2d.T)
actual_B_4d = np.reshape(actual_B.T, data.shape[:-1] + (-1, ))
exp_design, exp_B_4d, = glm(data, convolved)
assert_almost_equal(actual_B_4d, exp_B_4d)
def test_glm1():
actual_design = np.ones((len(convolved), 2))
actual_design[:, 1] = convolved
data_2d = np.reshape(data, (-1, data.shape[-1]))
actual_B = npl.pinv(actual_design).dot(data_2d.T)
actual_B_4d = np.reshape(actual_B.T, data.shape[:-1] + (-1,))
exp_design , exp_B_4d, = glm(data, convolved)
assert_almost_equal(actual_B_4d, exp_B_4d)
def gnl(g,m,alin=[],iterations=1,repeat=1,usei=False):
"""a function to just add some graph update functionality without relearning the graph, defined by m.graph*. Can use usei to use inverted Graph update layers instead of the normal ones (to make invertibility easier). Also understands alin (iarities) as a vector"""
if m.shallcomplex:
return gq(g,m,steps=m.complexsteps)
if usei:
g.X=glim(gs=g.s.gs,param=g.s.param,iterations=iterations,alinearity=alin,self_initializer=m.graph_init_self,neig_initializer=m.graph_init_neig)([g.A,g.X])
else:
g.X=glm(gs=g.s.gs,param=g.s.param,iterations=iterations,alinearity=alin,self_initializer=m.graph_init_self,neig_initializer=m.graph_init_self)([g.A,g.X])
if repeat>1:return gnl(g,alin=alin,iterations=iterations,repeat=repeat-1,usei=usei)
return g
def test_glm():
actual_design = np.ones((len(convolved), 2))
actual_design[:, 1] = convolved
exp_design, exp_B_4d = glm(data, convolved)
assert_almost_equal(actual_design, exp_design)
def test_glm():
actual_design = np.ones((len(convolved), 2))
actual_design[:, 1] = convolved
exp_design , exp_B_4d = glm(data, convolved)
assert_almost_equal(actual_design, exp_design)
