def test_det_net_set_diagonal_weights_rule(self):
network = pr.DeterministicHopfieldNetwork()
network.set_diagonal_weights_rule("zero")
self.assertTrue(network.diagonal_weights_equal_zero)
network.set_diagonal_weights_rule("non-zero")
self.assertFalse(network.diagonal_weights_equal_zero)
self.assertRaises(KeyError, network.set_diagonal_weights_rule,
*("asd", ))
def test_det_net_generate_weights_shape(self):
n_patterns = 10
n_bits = 5
patterns = utils.generate_n_random_patterns(n_patterns, n_bits)
network = pr.DeterministicHopfieldNetwork()
network.set_diagonal_weights_rule("zero")
network.set_patterns(patterns)
network.generate_weights()
self.assertEqual(network.weights.shape, (n_bits, n_bits))
def test_det_net_generate_weights_test_case_1(self):
patterns = np.array([[1, 2], [3, 4]])
network = pr.DeterministicHopfieldNetwork()
network.set_diagonal_weights_rule("non-zero")
network.set_patterns(patterns)
network.generate_weights()
expected_weights = np.array([[5, 7], [7, 10]])
expected = (network.weights == expected_weights).all()
self.assertTrue(expected)
def test_det_net_generate_weights_diagonal(self):
n_patterns = 6
n_bits = 100
patterns = utils.generate_n_random_patterns(n_patterns, n_bits)
network = pr.DeterministicHopfieldNetwork()
network.set_patterns(patterns)
network.set_diagonal_weights_rule("zero")
network.generate_weights()
diagonal = network.weights.diagonal()
self.assertFalse(np.any(diagonal))
def test_det_net_generate_weights_test_case_3(self):
patterns = np.array([[-1, 1, 1], [1, -1, 1]])
network = pr.DeterministicHopfieldNetwork()
network.set_diagonal_weights_rule("non-zero")
network.set_patterns(patterns)
network.generate_weights()
expected_weights = 1 / 3 * np.array([[2, -2, 0], [-2, 2, 0], [0, 0, 2]
])
expected = (network.weights == expected_weights).all()
self.assertTrue(expected)
def test_det_net_update_neuron_shape(self):
n_patterns = 5
n_bits = 100
patterns = utils.generate_n_random_patterns(n_patterns, n_bits)
network = pr.DeterministicHopfieldNetwork()
network.set_patterns(patterns)
network.set_diagonal_weights_rule("zero")
network.generate_weights()
original_pattern = patterns[3, :]
neuron_index = 10
updated_pattern = network.update_neuron(original_pattern, neuron_index)
self.assertEqual(updated_pattern.shape, original_pattern.shape)
def test_det_net_update_random_neuron(self):
n_patterns = 20
n_bits = 100
patterns = utils.generate_n_random_patterns(n_patterns, n_bits)
network = pr.DeterministicHopfieldNetwork()
network.set_patterns(patterns)
network.set_diagonal_weights_rule("zero")
network.generate_weights()
original_pattern = patterns[15, :]
updated_pattern = network.update_random_neuron(original_pattern)
n_differences = sum(updated_pattern != original_pattern)
self.assertTrue(n_differences <= 1)
def test_det_net_set_patterns(self):
patterns = utils.generate_n_random_patterns(3, 5)
network = pr.DeterministicHopfieldNetwork()
network.set_patterns(patterns)
expected = (network.patterns == patterns).all()
self.assertTrue(expected)
def test_det_net_get_state_of_local_field(self):
net = pr.DeterministicHopfieldNetwork()
self.assertEqual(net.get_state_of_local_field(0.1), 1)
self.assertEqual(net.get_state_of_local_field(0), 1)
self.assertEqual(net.get_state_of_local_field(-1), -1)
