def test_train_with_gradient_clipping(p_model, q_model):
train(p_model, q_model, cross_entropy, epochs=3, clip_gradients=1.)
train(p_model, q_model, perplexity, epochs=3, clip_gradients=1.)
train(p_model, q_model, forward_kl, epochs=3, clip_gradients=1.)
train(p_model, q_model, reverse_kl, epochs=3, clip_gradients=1.)
train(p_model, q_model, js_divergence, epochs=3, clip_gradients=1.)
def test_train(p_model, q_model):
train(p_model, q_model, cross_entropy, epochs=3)
train(p_model, q_model, perplexity, epochs=3)
train(p_model, q_model, forward_kl, epochs=3)
train(p_model, q_model, reverse_kl, epochs=3)
train(p_model, q_model, js_divergence, epochs=3)
def test_gan_train(gan):
if gan.n_dims == 1:
q_model = MixtureModel([Normal([-0.5],[[1.0]]), Normal([0.5],[[1.0]])], [0.5, 0.5])
p_model = MixtureModel([Normal([2.3], [[2.2]]), Normal([-2.3], [[2.2]])], [0.5, 0.5])
else:
q_model = MixtureModel([Normal([0., 0.], [1., 0., 0., 1.0]), Normal([0., 0.], [1., 0., 0., 1.0])], [0.25, 0.75])
p_model = MixtureModel([Normal([0., 0.], [1., 0., 0., 1.0]), Normal([0., 0.], [1., 0., 0., 1.0])], [0.25, 0.75])
train(p_model, q_model, gan, optimizer="RMSprop", epochs=3, lr=1e-3, batch_size=512)
X = p_model.sample(100)
gan.classify(X)
def fit(self, X, variational_dist=None, elbo_kwargs={}, **kwargs):
if variational_dist is None:
variational_dist = PPCA_Variational_V2(self)
data = Data(X)
stats = train(data, self, ELBO(variational_dist, **elbo_kwargs),
**kwargs)
return stats
def fit(self, x, **kwargs):
data = Data(x)
stats = train(data,
self.model,
self.criterion,
optimizer='RMSprop',
track_parameters=False,
**kwargs)
return stats
def test_plot_stats(p_model, q_model, n_stats):
stats = train(p_model,
q_model,
forward_kl,
epochs=10,
track_parameters=True)
assert len(stats.data['loss']) == 10
assert len(stats.data.keys()) == n_stats + 1
for key in stats.data.keys():
assert len(stats.data[key]) == 10
if isinstance(p_model, Normal):
plot_stats(stats, goals=[p_model.loc, p_model.scale])
elif isinstance(p_model, Exponential):
plot_stats(stats, goals=[p_model.rate])
elif isinstance(p_model, Beta):
plot_stats(stats, goals=[p_model.alpha, p_model.beta])
elif isinstance(p_model, StudentT):
plot_stats(stats, goals=[p_model.df, p_model.loc, p_model.scale])
else:
plot_stats(stats)
plt.close()
def fit(self, x, **kwargs):
data = Data(x)
stats = train(data, self.model, cross_entropy, **kwargs)
return stats
def fit(self, x, **kwargs):
data = Data(x)
return train(data, self, self.criterion, **kwargs)
def fit(self, R, **kwargs):
data = Data(R.view(-1, self.N * self.M))
stats = train(data, self, cross_entropy, **kwargs)
return stats
def test_train_with_penalty(p_model, q_model):
train(p_model, q_model, cross_entropy, epochs=3, l1_penalty=0.)
train(p_model, q_model, cross_entropy, epochs=3, l2_penalty=0.)
train(p_model, q_model, forward_kl, epochs=3, l1_penalty=0.)
train(p_model, q_model, forward_kl, epochs=3, l2_penalty=0.)
train(p_model,
q_model,
cross_entropy,
epochs=3,
l1_penalty=2.,
l2_penalty=2.)
train(p_model, q_model, forward_kl, epochs=3, l1_penalty=2., l2_penalty=2.)
def fit(self, x, use_elbo=True, **kwargs):
data = Data(x)
if use_elbo:
return train(data, self, self.criterion, **kwargs)
return train(data, self, cross_entropy, **kwargs)
