def __init__(self, dims):
super(Encoder, self).__init__()
[x_dim, h_dim, z_dim] = dims
self.linear = nn.Linear(x_dim, h_dim)
self.batch_norm = nn.BatchNorm1d(h_dim)
self.sample = GaussianSample(h_dim, z_dim)
def __init__(self, dims):
super(Encoder, self).__init__()
[x_dim, h_dim, z_dim] = dims
neurons = [x_dim, *h_dim]
linear_layers = [nn.Linear(neurons[i-1], neurons[i]) for i in range(1, len(neurons))]
self.hidden = nn.ModuleList(linear_layers)
self.sample = GaussianSample(h_dim[-1], z_dim)
def __init__(self, dims):
super(LadderEncoder, self).__init__()
[x_dim, h_dim, self.z_dim] = dims
self.in_features = x_dim
self.out_features = h_dim
self.linear = nn.Linear(x_dim, h_dim)
self.batchnorm = nn.BatchNorm1d(h_dim)
self.sample = GaussianSample(h_dim, self.z_dim)
def __init__(self, dims):
super(LadderDecoder, self).__init__()
[x_dim, h_dim, self.z_dim] = dims
self.linear1 = nn.Linear(x_dim, h_dim)
self.batchnorm1 = nn.BatchNorm1d(h_dim)
self.merge = GaussianMerge(h_dim, self.z_dim)
self.linear2 = nn.Linear(x_dim, h_dim)
self.batchnorm2 = nn.BatchNorm1d(h_dim)
self.sample = GaussianSample(h_dim, self.z_dim)
def __init__(self, dims):
super(Decoder, self).__init__()
[z_dim, h_dim, x_dim] = dims
# part 1
self.linear_1 = nn.Linear(x_dim, h_dim)
self.batch_norm_1 = nn.BatchNorm1d(h_dim)
self.merge = GaussianMerge(h_dim, z_dim)
# part 2
self.linear_2 = nn.Linear(x_dim, h_dim)
self.batch_norm_2 = nn.BatchNorm1d(h_dim)
self.sample = GaussianSample(h_dim, z_dim)
def __init__(self, dims):
"""
The ladder encoder differs from the standard encoder
by using batch-normalization and LReLU activation.
Additionally, it also returns the transformation x.
:param dims: dimensions [input_dim, [hidden_dims], [latent_dims]].
"""
super(LadderEncoder, self).__init__()
[x_dim, h_dim, self.z_dim] = dims
self.in_features = x_dim
self.out_features = h_dim
self.linear = nn.Linear(x_dim, h_dim)
self.batchnorm = nn.BatchNorm1d(h_dim)
self.sample = GaussianSample(h_dim, self.z_dim)
def __init__(self, dims):
"""
The ladder dencoder differs from the standard encoder
by using batch-normalization and LReLU activation.
Additionally, it also returns the transformation x.
:param dims: dimensions of the networks
given by the number of neurons on the form
[latent_dim, [hidden_dims], input_dim].
"""
super(LadderDecoder, self).__init__()
[self.z_dim, h_dim, x_dim] = dims
self.linear1 = nn.Linear(x_dim, h_dim)
self.batchnorm1 = nn.BatchNorm1d(h_dim)
self.merge = GaussianMerge(h_dim, self.z_dim)
self.linear2 = nn.Linear(x_dim, h_dim)
self.batchnorm2 = nn.BatchNorm1d(h_dim)
self.sample = GaussianSample(h_dim, self.z_dim)
def __init__(self, dims):
"""
Inference network
Attempts to infer the probability distribution
p(z|x) from the data by fitting a variational
distribution q_φ(z|x). Returns the two parameters
of the distribution (µ, log σ²).
:param dims: dimensions of the networks
given by the number of neurons on the form
[input_dim, [hidden_dims], latent_dim].
"""
super(Encoder, self).__init__()
[x_dim, h_dim, z_dim] = dims
neurons = [x_dim, *h_dim]
linear_layers = [
nn.Linear(neurons[i - 1], neurons[i])
for i in range(1, len(neurons))
]
self.hidden = nn.ModuleList(linear_layers)
self.sample = GaussianSample(h_dim[-1], z_dim)