def activate(self, input_list):
if (len(input_list) != len(self.hpneat_config.input_neuron_position)):
raise ValueError("argument ouf of range")
pre_activate_val = copy.deepcopy(self.activate_val)
for n in range(len(input_list)):
self.activate_val[self.hpneat_config.input_neuron_position[n][0]][
self.hpneat_config.input_neuron_position[n][1]] = input_list[n]
activate_vec = tools.matrix_to_vector(self.activate_val)
activate_vec = np.dot(activate_vec, self.weight)
self.activate_val = tools.sigmoid_for_np_ndarray(
tools.vector_to_matrix(activate_vec) + self.bias)
#update weight
self.weight = update_weight_for_abc(self.hpneat_config.num_x, \
self.hpneat_config.num_y, \
self.weight, \
pre_activate_val, \
self.activate_val,\
self.A, self.B, self.C, self.ita)
output_vec = []
for n in range(len(self.hpneat_config.output_neuron_position)):
output_vec.append(self.activate_val[
self.hpneat_config.output_neuron_position[n][0]][
self.hpneat_config.output_neuron_position[n][1]])
return output_vec
def test_matrix_to_vector():
hpneat_config.num_x = 2
hpneat_config.num_y = 3
m1 = 20 * np.random.random((hpneat_config.num_x, hpneat_config.num_y)) - 10
print(m1)
vec = tools.matrix_to_vector(m1)
print(vec)
print(vec.shape)
m2 = tools.vector_to_matrix(vec)
print(m2)
assert m1.shape == (2, 3)
assert vec.shape == (1, 6)
assert m1[1][0] == m2[1][0]
assert m1.shape == m2.shape
def testWolfBQ():
r = 0.1327
T = np.linspace(0, 1, 100)
BQ1 = []
BQ2 = []
for i, t in enumerate(T):
print(f'{i+1}/{len(T)}')
B = functional(t, r)
bq1, _, _, _, _, _, _ = CHSH.chsh2(T.vector_to_matrix(B), 0.0001)
bq2 = QValue(t, r)
BQ1.append(bq1)
BQ2.append(bq2)
plt.plot(BQ1)
plt.plot(BQ2)
plt.show()
def activate(self, input_list):
if (len(input_list) != len(self.hpneat_config.input_neuron_position)):
raise ValueError("argument ouf of range")
for n in range(len(input_list)):
self.activate_val[self.hpneat_config.input_neuron_position[n][0]][
self.hpneat_config.input_neuron_position[n][1]] = input_list[n]
activate_vec = tools.matrix_to_vector(self.activate_val)
activate_vec = np.dot(activate_vec, self.weight)
self.activate_val = tools.sigmoid_for_np_ndarray(
tools.vector_to_matrix(activate_vec) + self.bias)
output_vec = []
for n in range(len(self.hpneat_config.output_neuron_position)):
output_vec.append(self.activate_val[
self.hpneat_config.output_neuron_position[n][0]][
self.hpneat_config.output_neuron_position[n][1]])
return output_vec