def __init__(self,
num_users,
num_items,
embedding_dim=32,
user_embedding_layer=None,
item_embedding_layer=None,
sparse=False):
super(BilinearNet, self).__init__()
self.embedding_dim = embedding_dim
if user_embedding_layer is not None:
self.user_embeddings = user_embedding_layer
else:
self.user_embeddings = ScaledEmbedding(num_users,
embedding_dim,
sparse=sparse)
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
sparse=sparse)
self.user_biases = ZeroEmbedding(num_users, 1, sparse=sparse)
self.item_biases = ZeroEmbedding(num_items, 1, sparse=sparse)
def __init__(self, num_users, num_items, embedding_dim=32, sparse=False):
super().__init__()
self.embedding_dim = embedding_dim
self.user_embeddings = ScaledEmbedding(num_users, embedding_dim,
sparse=sparse)
self.item_embeddings = ScaledEmbedding(num_items, embedding_dim,
sparse=sparse)
self.user_biases = ZeroEmbedding(num_users, 1, sparse=sparse)
self.item_biases = ZeroEmbedding(num_items, 1, sparse=sparse)
def __init__(self,
num_users,
num_items,
embedding_dim=32,
user_embedding_layer=None,
item_embedding_layer=None,
sparse=False):
super(RankingNet, self).__init__()
self.embedding_dim = embedding_dim
# self.inputDim = embedding_dim * 3
# self.hiddenDim = self.inputDim * 2
# self.hiddenDim2 = embedding_dim * 3
# self.l2Output = 1
self.inputDim = embedding_dim * 2
self.hiddenDim = self.inputDim * 1
self.hiddenDim2 = embedding_dim * 2
self.l2Output = 1
if user_embedding_layer is not None:
self.user_embeddings = user_embedding_layer
else:
self.user_embeddings = ScaledEmbedding(num_users,
embedding_dim,
sparse=sparse)
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
sparse=sparse)
self.user_biases = ZeroEmbedding(num_users, 1, sparse=sparse)
self.item_biases = ZeroEmbedding(num_items, 1, sparse=sparse)
# self.linear1 = nn.Linear(self.inputDim, self.hiddenDim)
# self.bn1 = nn.BatchNorm1d(self.hiddenDim)
# self.linear2 = nn.Linear(self.hiddenDim, self.l2Output)
# self.dropout = nn.Dropout(p = 0.8)
# self.output = nn.Sigmoid()
self.linear1 = nn.Linear(self.inputDim, self.hiddenDim)
self.bn1 = nn.BatchNorm1d(self.hiddenDim)
self.linear2 = nn.Linear(self.hiddenDim, self.hiddenDim2)
self.bn2 = nn.BatchNorm1d(self.hiddenDim2)
self.linear3 = nn.Linear(self.hiddenDim2, self.l2Output)
self.dropout = nn.Dropout(p=0.8)
self.output = nn.Sigmoid()
def __init__(self, num_users, num_items, embedding_dim=32,
num_components=4):
super(EmbeddingMixtureNet, self).__init__()
self.embedding_dim = embedding_dim
self.num_components = num_components
self.taste_embeddings = ScaledEmbedding(num_users, embedding_dim * num_components)
self.attention_embeddings = ScaledEmbedding(num_users, embedding_dim * num_components)
self.item_embeddings = ScaledEmbedding(num_items, embedding_dim)
self.user_biases = ZeroEmbedding(num_users, 1)
self.item_biases = ZeroEmbedding(num_items, 1)
def __init__(self, num_users, num_items, embedding_dim=32,
num_components=4):
super(NonlinearMixtureNet, self).__init__()
self.embedding_dim = embedding_dim
self.num_components = num_components
self.user_embeddings = nn.Embedding(num_users, embedding_dim)
self.item_embeddings = nn.Embedding(num_items, embedding_dim)
self.user_biases = ZeroEmbedding(num_users, 1)
self.item_biases = ZeroEmbedding(num_items, 1)
self.mixture = MixtureComponent(embedding_dim, num_components)
def __init__(self, num_items, embedding_dim=32,
num_components=4,
diversity_penalty=1.0,
item_embedding_layer=None, sparse=False):
super(DiversifiedMixtureLSTMNet, self).__init__()
self.embedding_dim = embedding_dim
self.num_components = num_components
self._diversity_penalty = diversity_penalty
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items, embedding_dim,
padding_idx=PADDING_IDX,
sparse=sparse)
self.item_biases = ZeroEmbedding(num_items, 1, sparse=sparse,
padding_idx=PADDING_IDX)
self.lstm = nn.LSTM(batch_first=True,
input_size=embedding_dim,
hidden_size=embedding_dim)
self.projection = nn.Conv1d(embedding_dim,
embedding_dim * self.num_components * 2,
kernel_size=1)
def __init__(self,
num_items,
embedding_dim=32,
item_embedding_layer=None,
sparse=False):
super(LSTMNet, self).__init__()
self.embedding_dim = embedding_dim
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
padding_idx=PADDING_IDX,
sparse=sparse)
self.item_biases = ZeroEmbedding(num_items,
1,
sparse=sparse,
padding_idx=PADDING_IDX)
self.lstm = nn.LSTM(batch_first=True,
input_size=embedding_dim,
hidden_size=embedding_dim)
def __init__(self,
num_items,
embedding_dim=32,
kernel_width=5,
dilation=1,
num_layers=1,
sparse=False):
super().__init__()
self.embedding_dim = embedding_dim
self.kernel_width = kernel_width
self.dilation = dilation
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
sparse=sparse,
padding_idx=PADDING_IDX)
self.item_biases = ZeroEmbedding(num_items,
1,
sparse=sparse,
padding_idx=PADDING_IDX)
self.cnn_layers = [
nn.Conv2d(embedding_dim,
embedding_dim, (kernel_width, 1),
dilation=(dilation, 1)) for _ in range(num_layers)
]
for i, layer in enumerate(self.cnn_layers):
self.add_module('cnn_{}'.format(i), layer)
def __init__(self, num_items, embedding_dim=32, sparse=False):
super(PoolNet, self).__init__()
self.embedding_dim = embedding_dim
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
sparse=sparse,
padding_idx=PADDING_IDX)
self.item_biases = ZeroEmbedding(num_items,
1,
sparse=sparse,
padding_idx=PADDING_IDX)
def __init__(self,
num_items,
embedding_dim=32,
kernel_width=3,
dilation=1,
num_layers=1,
nonlinearity='tanh',
residual_connections=True,
sparse=False,
benchmark=True,
item_embedding_layer=None):
super(CNNNet, self).__init__()
cudnn.benchmark = benchmark
self.embedding_dim = embedding_dim
self.kernel_width = _to_iterable(kernel_width, num_layers)
self.dilation = _to_iterable(dilation, num_layers)
if nonlinearity == 'tanh':
self.nonlinearity = F.tanh
elif nonlinearity == 'relu':
self.nonlinearity = F.relu
else:
raise ValueError('Nonlinearity must be one of (tanh, relu)')
self.residual_connections = residual_connections
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
padding_idx=PADDING_IDX,
sparse=sparse)
self.item_biases = ZeroEmbedding(num_items,
1,
sparse=sparse,
padding_idx=PADDING_IDX)
self.cnn_layers = [
nn.Conv2d(embedding_dim,
embedding_dim, (_kernel_width, 1),
dilation=(_dilation, 1))
for (_kernel_width,
_dilation) in zip(self.kernel_width, self.dilation)
]
for i, layer in enumerate(self.cnn_layers):
self.add_module('cnn_{}'.format(i), layer)
def __init__(self, num_users, num_items, embedding_dim=32,
projection_scale=1.0,
num_components=4):
super(MixtureNet, self).__init__()
self.embedding_dim = embedding_dim
self.num_components = num_components
self.projection_scale = projection_scale
self.user_embeddings = ScaledEmbedding(num_users, embedding_dim)
self.item_embeddings = ScaledEmbedding(num_items, embedding_dim)
self.user_biases = ZeroEmbedding(num_users, 1)
self.item_biases = ZeroEmbedding(num_items, 1)
self.taste_projection = nn.Linear(embedding_dim,
embedding_dim * self.num_components, bias=False)
self.attention_projection = nn.Linear(embedding_dim,
embedding_dim * self.num_components, bias=False)
for layer in (self.taste_projection, self.attention_projection):
torch.nn.init.xavier_normal(layer.weight, self.projection_scale)
def __init__(self,
num_users,
num_items,
embedding_dim=32,
n_layers=2,
sparse=None):
super(DeepNet, self).__init__()
self.emb_dim = embedding_dim
#self.user_emb = nn.Embedding(num_users, self.emb_dim)
#self.item_emb = nn.Embedding(num_items, self.emb_dim)
self.user_emb = ScaledEmbedding(num_users,
embedding_dim,
sparse=sparse)
self.item_emb = ScaledEmbedding(num_items,
embedding_dim,
sparse=sparse)
self.mlp = nn.Sequential(nn.Linear(self.emb_dim, 16), nn.ReLU(),
nn.Dropout(p=0.8), nn.Linear(16, 32),
nn.ReLU(), nn.Dropout(0.8))
self.user_biases = ZeroEmbedding(num_users, 1)
self.item_biases = ZeroEmbedding(num_items, 1)
self.out = nn.ReLU()
def __init__(self, num_items, embedding_dim=32, item_embedding_layer=None):
super().__init__()
self.embedding_dim = embedding_dim
self.item_biases = ZeroEmbedding(num_items, 1, padding_idx=PADDING_IDX)
if item_embedding_layer is not None:
self.item_embedding = item_embedding_layer
else:
self.item_embedding = ScaledEmbedding(num_items,
embedding_dim,
padding_idx=PADDING_IDX)
# why input_size and hidden_size needs to be both embedding_dim ??
self.lstm = nn.LSTM(input_size=embedding_dim,
hidden_size=embedding_dim,
batch_first=True)
def __init__(self, num_items, embedding_dim=32,
item_embedding_layer=None, sparse=False):
super(PoolNet, self).__init__()
self.embedding_dim = embedding_dim
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items, embedding_dim,
padding_idx=PADDING_IDX,
sparse=sparse)
self.item_biases = ZeroEmbedding(num_items, 1, sparse=sparse,
padding_idx=PADDING_IDX)
def __init__(self, num_items, embedding_dim=32, sparse=False):
super().__init__()
self.embedding_dim = embedding_dim
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
sparse=sparse,
padding_idx=PADDING_IDX)
self.item_biases = ZeroEmbedding(num_items,
1,
sparse=sparse,
padding_idx=PADDING_IDX)
self.lstm = nn.LSTM(batch_first=True,
input_size=embedding_dim,
hidden_size=embedding_dim)
def __init__(self,
num_items,
embedding_dim=32,
item_embedding_layer=None,
sparse=False,
layers=[],
nonlinearity='tanh'):
super(PoolNet, self).__init__()
self.embedding_dim = embedding_dim
if nonlinearity == 'tanh':
self.nonlinearity = F.tanh
elif nonlinearity == 'relu':
self.nonlinearity = F.relu
else:
raise ValueError('Nonlinearity must be one of (tanh, relu)')
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items,
embedding_dim,
padding_idx=PADDING_IDX,
sparse=sparse)
self.item_biases = ZeroEmbedding(num_items,
1,
sparse=sparse,
padding_idx=PADDING_IDX)
self.fc_layers = None
if layers and len(layers) > 0:
self.fc_layers = torch.nn.ModuleList()
for idx, (in_size,
out_size) in enumerate(zip(layers[:-1], layers[1:])):
self.fc_layers.append(
TimeDistributed(torch.nn.Linear(in_size, out_size)))
def __init__(self, num_items, embedding_dim=32,
item_embedding_layer=None, sparse=False):
super(CustomPoolNet, self).__init__()
self.embedding_dim = embedding_dim
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
# Return embedding vector for each item, size 3
# Examples: embedding([1,2,4,5])
# tensor(
# [[[-0.0251, -1.6902, 0.7172],
# [-0.6431, 0.0748, 0.6969],
# [ 1.4970, 1.3448, -0.9685],
# [-0.3677, -2.7265, -0.1685]]])
self.item_embeddings = ScaledEmbedding(num_items, embedding_dim,
padding_idx=PADDING_IDX,
sparse=sparse)
self.item_biases = ZeroEmbedding(num_items, 1, sparse=sparse,
padding_idx=PADDING_IDX)
def __init__(self, num_items, embedding_dim=32,
item_embedding_layer=None, sparse=False):
super(mLSTMNet, self).__init__()
self.embedding_dim = embedding_dim
if item_embedding_layer is not None:
self.item_embeddings = item_embedding_layer
else:
self.item_embeddings = ScaledEmbedding(num_items, embedding_dim,
padding_idx=PADDING_IDX,
sparse=sparse)
self.item_biases = ZeroEmbedding(num_items, 1, sparse=sparse,
padding_idx=PADDING_IDX)
h_init = torch.zeros(embedding_dim)
h_init.normal_(0, 1.0 / self.embedding_dim)
self.h_init = nn.Parameter(h_init, requires_grad=True)
self.mlstm = mLSTM(input_size=embedding_dim,
hidden_size=embedding_dim)
