def test_generate_train_max_weirdshapes(self):
mz1 = Maze((3, 4, 2, 2))
mz1.set_spike_train_params(2)
spikes = mz1.generate_spike_train((2, 3, 0, 1))
self.assertEqual(spikes.shape, (2, 3, 4, 2, 2))
mz2 = Maze(10)
mz2.set_spike_train_params(200)
spikes = mz2.generate_spike_train([5])
self.assertEqual(spikes.shape, (200, 10))
def test_generate_possible_moves(self):
mz = Maze(size=(4, 7))
d = mz._generate_possible_moves(dims=4, symmetric=True)
self.assertEqual(len(d.keys()), 8)
self.assert_(np.array_equal(d[0], np.array([1,0])))
self.assert_(np.array_equal(d[1], np.array([3,0])))
self.assert_(np.array_equal(d[2], np.array([0,1])))
self.assert_(np.array_equal(d[3], np.array([0,6])))
for i in range(100):
d = mz._generate_possible_moves(dims=4, symmetric=True)
self.assertLess(d[4][0], 4)
self.assertLess(d[4][1], 7)
self.assertGreater(d[4][0], -4)
self.assertGreater(d[4][1], -7)
self.assert_(np.array_equal(d[5], -d[4]))
d = mz._generate_possible_moves(dims=1, symmetric=True)
self.assertEqual(len(d.keys()), 2)
self.assert_(np.array_equal(d[0], np.array([1, 0])))
self.assert_(np.array_equal(d[1], np.array([3, 0])))
d = mz._generate_possible_moves(dims=16, symmetric=False)
self.assertEqual(len(d.keys()), 16)
def test_get_random_position(self):
mz = Maze(size=(5, 6, 7))
posz = []
for i in range(100):
posz.append(mz.get_random_position())
posz = np.array(posz)
self.assertLess(posz.max(axis=0)[0], 5)
self.assertLess(posz.max(axis=0)[1], 6)
self.assertLess(posz.max(axis=0)[2], 7)
mz = Maze(size=12)
posz = []
for i in range(100):
posz.append(mz.get_random_position())
posz = np.array(posz)
self.assertLess(posz.max(axis=0), 12)
def test_init(self):
mz = Maze(size=5)
self.assertEqual(mz.maze_dimensionality, 1)
self.assertEqual(mz.current_pos.shape, (1,))
self.assertEqual(mz.current_pos[0].shape, ())
def test_generate_train_shape(self):
mz = Maze((3, 4))
mz.set_spike_train_params(20)
spikes = mz.generate_spike_train((1, 1))
self.assertEqual(spikes.shape, (20, 3, 4))
def test_generate_train_max_spikes(self):
mz = Maze((3, 4))
mz.set_spike_train_params(20)
spikes = mz.generate_spike_train((1, 1))
self.assertEqual(spikes[:, 1, 1].sum(), 20)
theta = 0.3169981156555285
eta1 = 0.0009796194970225012
eta2 = 0.015360662582262801
dims = 2
seed = 635
tf.random.set_seed(seed)
rd.seed(seed)
t_steps = 100
epochs = 1500 * 6
out_in_ratio = 2**3
# Initialize maze
mz = Maze((4, 4))
t_mat = mz.build_transition_matrix(dims=dims, symmetric=True)
mz.set_spike_train_params(t_steps, high_freq_p=1.0, low_freq_p=0.0)
n_in = mz.tot_room_number
n_pop = mz.tot_room_number * out_in_ratio
dt = 1
logdir = "logs/" + f"{n_in}+{n_pop}/" + "last/" + "d2/" + "thresh_norm/" + f"seed{seed}/" + datetime.now(
).strftime("%m%d-%H%M")
writer = tf.summary.create_file_writer(logdir)
# Initialize network [two input sets of size n_in and two output populations of size n_pop]
# Block identity matrix
obs_m1 = (1.0) * np.kron(np.eye(n_in), np.ones([1, out_in_ratio
])) #martin put them to 1 too