def test_mu(self):
""" check mu given by get_conditional_negative_energy
"""
in_patterns = [
amath.array([1, 0, 0, 0]),
amath.array([0, 1, 0, 0]),
amath.array([0, 0, 1, 0])
]
self.model._update_state(in_patterns[2])
self.model._update_state(in_patterns[1])
self.model._update_state(in_patterns[0])
fifo_array = self.model.fifo.to_array()
self.assertTrue(amath.allclose(fifo_array[0], in_patterns[0]))
self.assertTrue(amath.allclose(fifo_array[1], in_patterns[1]))
self.assertTrue(amath.allclose(fifo_array[2], in_patterns[2]))
mu_test = deepcopy(self.model.variables["b"])
u_tilde = self.model._get_u_tilde()
# naive implementation of sigmoid
sig_u_tilde = sigmoid(u_tilde)
for d in xrange(self.order):
mu_test += self.model.variables["W"][d, d, :]
mu_test += sig_u_tilde.dot(self.model.variables["V"])
self.assertTrue(
amath.allclose(self.model._get_conditional_negative_energy(),
mu_test))
pass
def testFifo(self):
"""
testing fifo and _update_state method in VectorRegressionWithVariance
"""
print("\n * testing fifo and update_state method "
"in VectorRegressionWithVariance \n")
in_dim = 3
order = 3
len_ts = 10
model = VectorRegressionWithVariance(in_dim, in_dim, order)
random = amath.random.RandomState(0)
in_patterns = amath.random.uniform(size=(len_ts, in_dim))
fifo_test = amath.zeros((order, in_dim))
for i in xrange(len_ts):
self.assertTrue(amath.allclose(model.fifo.to_array(), fifo_test))
model.learn_one_step(in_patterns[i])
popped_in_pattern = model._update_state(in_patterns[i])
if i < order:
self.assertTrue(
amath.allclose(popped_in_pattern, amath.zeros(in_dim)))
else:
self.assertTrue(
amath.allclose(popped_in_pattern, in_patterns[i - order]))
fifo_test[1:] = fifo_test[:-1]
fifo_test[0] = in_patterns[i]
def testFifo(self):
"""
testing fifo and _update_state method in MultiTargetVectorRegression
"""
print("\n * testing fifo and update_state method "
"in MultiTargetVectorRegression \n")
in_dim = 3
out_dims = [2, 4]
SGDs = [AdaGrad(), AdaGrad()]
order = 3
len_ts = 10
model = MultiTargetVectorRegression(in_dim, out_dims, SGDs, order)
random = amath.random.RandomState(0)
in_patterns = amath.random.uniform(size=(len_ts, in_dim))
out_pattern_0 = amath.random.uniform(size=(len_ts, out_dims[0]))
out_pattern_1 = amath.random.uniform(size=(len_ts, out_dims[1]))
fifo_test = amath.zeros((order, in_dim))
for i in xrange(len_ts):
self.assertTrue(
amath.allclose(model.layers[0].fifo.to_array(), fifo_test))
model.learn_one_step([out_pattern_0[i], out_pattern_1[i]])
popped_in_pattern = model._update_state(in_patterns[i])
if i < order:
self.assertTrue(
amath.allclose(popped_in_pattern, amath.zeros(in_dim)))
else:
self.assertTrue(
amath.allclose(popped_in_pattern, in_patterns[i - order]))
fifo_test[1:] = fifo_test[:-1]
fifo_test[0] = in_patterns[i]
def test_update_state(self):
print("""
---------------------------------
test updating internal states
---------------------------------
""")
anc_test = amath.random.uniform(self.low, self.high,
(self.n_anc, self.dim))
model = FunctionalDyBM(
self.dim, anc_test, self.delay, self.decay_rates)
loc = amath.random.uniform(self.low, self.high, (self.n_obs, self.dim))
for i in range(10):
pattern = amath.zeros(self.n_obs)
pattern[i % self.n_obs] = 1
fifo_old = model.fifo[:-1]
fifo_last = model.fifo[-1]
e_trace_old = model.e_trace
model._update_state(pattern, loc)
# check fifo update
self.assertTrue(amath.allclose(model.fifo[1:], fifo_old))
# check e_trace update
e_trace_new = amath.zeros((self.n_elg, self.n_anc)) + fifo_last
for k in xrange(e_trace_new.shape[0]):
e_trace_new[k, :] = e_trace_new[k, :] \
+ self.decay_rates[k] * e_trace_old[k, :]
self.assertTrue(amath.allclose(model.e_trace, e_trace_new))
# check mu_anc update
self.assertTrue(amath.allclose(model.mu, amath.zeros(self.n_anc)))
print("mu(P) = {}".format(model.mu))
return 0
def test_init(self):
print("""
---------------------------------------------
test initialization of NeuralFieldDyBM
---------------------------------------------
""")
anc_test = amath.random.uniform(self.low, self.high,
(self.n_anc, self.dim))
model = FunctionalDyBM(
self.dim, anc_test, self.delay, self.decay_rates)
# check anchors
amath.random.seed(self.rand_seed)
self.assertTrue(amath.allclose(model.anc_basis, anc_test))
# check the shapes
self.assertTrue(model.variables["W"].shape ==
(self.delay - 1, self.n_anc, self.n_anc))
self.assertTrue(model.variables["V"].shape ==
(self.n_elg, self.n_anc, self.n_anc))
self.assertTrue(model.variables["b"].shape == (self.n_anc,))
# check kernel
X1 = amath.random.randn(10, self.dim)
X2 = amath.random.randn(10, self.dim)
rbf = amath.kernel_metrics['rbf'](X1, X2, gamma=1.0)
self.assertTrue(amath.allclose(model.ker(X1, X2), rbf))
# check anchor points
self.assertTrue(model.anc_basis.shape == (self.n_anc, self.dim))
# check fifo and e_trace
self.assertTrue(model.fifo.shape == (self.delay - 1, self.n_anc))
self.assertTrue(model.e_trace.shape == (self.n_elg, self.n_anc))
return 0
def test_u_tilde(self):
""" check _get_u_tilde
"""
in_patterns = [amath.array([1, 0, 0, 0]),
amath.array([0, 1, 0, 0]),
amath.array([0, 0, 1, 0])]
self.model._update_state(in_patterns[2])
self.model._update_state(in_patterns[1])
self.model._update_state(in_patterns[0])
u_tilde_test = amath.zeros(self.dim_hidden)
for d in xrange(self.order):
u_tilde_test += self.model.variables["U"][d, d, :]
u_tilde_test += self.model.variables["b_h"]
self.assertTrue(amath.allclose(u_tilde_test, self.model._get_u_tilde()))
pass