def __init__(self, opt):
super(FM, self).__init__()
self.use_cuda = opt.get('use_cuda')
self.latent_dim = opt['latent_dim']
self.field_dims = opt['field_dims']
self.feature_num = sum(self.field_dims)
self.embedding = PEPEmbedding(opt)
self.linear = FeaturesLinear(self.field_dims) # linear part
self.fm = FactorizationMachine(reduce_sum=True)
print("BackBone Embedding Parameters: ", self.feature_num * self.latent_dim)
def __init__(self, field_dims, embed_dim):
super().__init__()
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.linear = FeaturesLinear(field_dims)
self.fm = FactorizationMachine(reduce_sum=True)
def __init__(self,
field_dims,
embed_dim,
atten_embed_dim,
num_heads,
num_layers,
mlp_dims,
dropouts,
has_residual=True):
super().__init__()
self.num_fields = len(field_dims)
self.linear = FeaturesLinear(field_dims)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.atten_embedding = torch.nn.Linear(embed_dim, atten_embed_dim)
self.embed_output_dim = len(field_dims) * embed_dim
self.atten_output_dim = len(field_dims) * atten_embed_dim
self.has_residual = has_residual
self.mlp = MultiLayerPerceptron(self.embed_output_dim, mlp_dims,
dropouts[1])
self.self_attns = torch.nn.ModuleList([
torch.nn.MultiheadAttention(atten_embed_dim,
num_heads,
dropout=dropouts[0])
for _ in range(num_layers)
])
self.attn_fc = torch.nn.Linear(self.atten_output_dim, 1)
if self.has_residual:
self.V_res_embedding = torch.nn.Linear(embed_dim, atten_embed_dim)
def __init__(self, field_dims, embed_dim, mlp_dims, dropout):
super().__init__()
self.linear = FeaturesLinear(field_dims)
self.fm = FactorizationMachine(reduce_sum=True)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.embed_output_dim = len(field_dims) * embed_dim
self.mlp = MultiLayerPerceptron(self.embed_output_dim, mlp_dims, dropout)
def __init__(self, field_dims, embed_dim, mlp_dims, dropouts):
super().__init__()
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.linear = FeaturesLinear(field_dims)
self.fm = torch.nn.Sequential(FactorizationMachine(reduce_sum=False),
torch.nn.BatchNorm1d(embed_dim),
torch.nn.Dropout(dropouts[0]))
self.mlp = MultiLayerPerceptron(embed_dim, mlp_dims, dropouts[1])
def __init__(self, field_dims, embed_dim, num_layers, mlp_dims, dropout):
super().__init__()
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.linear = FeaturesLinear(field_dims)
self.embed_output_dim = len(field_dims) * embed_dim
self.cn = CrossNetwork(self.embed_output_dim, num_layers)
self.cn_output = torch.nn.Linear(self.embed_output_dim, 1)
self.mlp = MultiLayerPerceptron(self.embed_output_dim, mlp_dims, dropout)
def __init__(self, field_dims, embed_dim, mlp_dims, dropout):
super().__init__()
self.linear = FeaturesLinear(field_dims)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.embed_output_dim = len(field_dims) * embed_dim
self.mlp = MultiLayerPerceptron(self.embed_output_dim, mlp_dims,
dropout)
def __init__(self, field_dims, embed_dim, attn_size, dropouts):
super().__init__()
self.num_fields = len(field_dims)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.linear = FeaturesLinear(field_dims)
self.afm = AttentionalFactorizationMachine(embed_dim, attn_size, dropouts)
def __init__(self, field_dims, embed_dim, LNN_dim, mlp_dims, dropouts):
super().__init__()
self.num_fields = len(field_dims)
self.linear = FeaturesLinear(field_dims) # Linear
self.embedding = FeaturesEmbedding(field_dims, embed_dim) # Embedding
self.LNN_dim = LNN_dim
self.LNN_output_dim = self.LNN_dim * embed_dim
self.LNN = LNN(self.num_fields, embed_dim, LNN_dim)
self.mlp = MultiLayerPerceptron(self.LNN_output_dim, mlp_dims,
dropouts[0])
def __init__(self, field_dims, embed_dim, mlp_dims, dropouts):
super().__init__()
self.linear = FeaturesLinear(field_dims)
self.ffm = FieldAwareFactorizationMachine(field_dims, embed_dim)
self.ffm_output_dim = len(field_dims) * (len(field_dims) -
1) // 2 * embed_dim
self.bn = torch.nn.BatchNorm1d(self.ffm_output_dim)
self.dropout = torch.nn.Dropout(dropouts[0])
self.mlp = MultiLayerPerceptron(self.ffm_output_dim, mlp_dims,
dropouts[1])
def __init__(self, field_dims, embed_dim, mlp_dims, dropout, method='inner'):
super().__init__()
num_fields = len(field_dims)
if method == 'inner':
self.pn = InnerProductNetwork()
elif method == 'outer':
self.pn = OuterProductNetwork(num_fields, embed_dim)
else:
raise ValueError('unknown product type: ' + method)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.linear = FeaturesLinear(field_dims, embed_dim)
self.embed_output_dim = num_fields * embed_dim
self.mlp = MultiLayerPerceptron(num_fields * (num_fields - 1) // 2 + self.embed_output_dim, mlp_dims, dropout)
def __init__(self, field_dims, obsItem_dims, obsUser_dims, embed_dim, obs,
embed):
super().__init__()
# print(field_dims, embed_dim)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
# print(field_dims[0:1], obsItem_dims)
# input()
# if obs:
self.obsItem_coeff = FeaturesEmbedding(field_dims[0:1], obsItem_dims)
self.obsUser_coeff = FeaturesEmbedding(field_dims[1:2], obsUser_dims)
self.linear = FeaturesLinear(field_dims, bias=False)
self.fm = FactorizationMachine(reduce_sum=True)
self.obs = obs
self.embed = embed
assert obs or embed, "One of obs or embed must be true\n"
def __init__(self,
field_dims,
embed_dim,
mlp_dims,
dropout,
cross_layer_sizes,
split_half=True):
super().__init__()
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.embed_output_dim = len(field_dims) * embed_dim
self.cin = CompressedInteractionNetwork(len(field_dims),
cross_layer_sizes, split_half)
self.mlp = MultiLayerPerceptron(self.embed_output_dim, mlp_dims,
dropout)
self.linear = FeaturesLinear(field_dims)
def __init__(self, field_dims, order, embed_dim):
super().__init__()
if order < 1:
raise ValueError(f'invalid order: {order}')
self.order = order
self.embed_dim = embed_dim
self.linear = FeaturesLinear(field_dims)
if order >= 2:
self.embedding = FeaturesEmbedding(field_dims,
embed_dim * (order - 1))
self.fm = FactorizationMachine(reduce_sum=True)
if order >= 3:
self.kernels = torch.nn.ModuleList([
AnovaKernel(order=i, reduce_sum=True)
for i in range(3, order + 1)
])
def __init__(self,
field_dims,
embed_dim,
mlp_dims,
dropout,
training_method='dfa'):
super().__init__()
self.linear = FeaturesLinear(field_dims)
self.fm = FactorizationMachine(reduce_sum=True)
self.embedding = FeaturesEmbedding(
field_dims, embed_dim) # Trained through FM. OK: no weights in FM.
self.embed_output_dim = len(field_dims) * embed_dim
self.mlp = DFAMultiLayerPerceptron(self.embed_output_dim,
mlp_dims,
dropout,
training_method=training_method)
def __init__(self, field_dims, embed_dim, num_heads, num_layers, mlp_dims,
dropouts):
super().__init__()
self.embed_dim = embed_dim
self.num_fields = len(field_dims)
self.linear = FeaturesLinear(field_dims)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.embed_output_dim = len(field_dims) * embed_dim
# self.res = torch.nn.Linear(self.embed_output_dim,self.embed_output_dim)
self.mlp = MultiLayerPerceptron(self.embed_output_dim + 399, mlp_dims,
dropouts[1])
self.self_attns = torch.nn.ModuleList([
torch.nn.MultiheadAttention(embed_dim,
num_heads,
dropout=dropouts[0])
for _ in range(num_layers)
])
self.attn_fc = torch.nn.Linear(self.embed_output_dim, 1)
def __init__(self,
field_dims,
embed_dim,
atten_embed_dim,
num_heads,
num_layers,
mlp_dims,
dropouts,
has_residual=True,
training_method='dfa'):
super().__init__()
self.num_fields = len(field_dims)
self.linear = FeaturesLinear(field_dims)
self.embedding = FeaturesEmbedding(field_dims, embed_dim)
self.dfa_embedding = DFALayer()
self.atten_embedding = torch.nn.Linear(embed_dim, atten_embed_dim)
self.dfa_atten_embedding = DFALayer()
self.embed_output_dim = len(field_dims) * embed_dim
self.atten_output_dim = len(field_dims) * atten_embed_dim
self.has_residual = has_residual
self.mlp = DFAMultiLayerPerceptron(self.embed_output_dim,
mlp_dims,
dropouts[1],
dfa_output=False)
self.self_attns = torch.nn.ModuleList([
torch.nn.MultiheadAttention(atten_embed_dim,
num_heads,
dropout=dropouts[0])
for _ in range(num_layers)
])
self.dfa_self_attns = [DFALayer() for _ in range(num_layers - 1)]
self.attn_fc = torch.nn.Linear(self.atten_output_dim, 1)
if self.has_residual:
self.V_res_embedding = torch.nn.Linear(embed_dim, atten_embed_dim)
self.dfa_cross = DFALayer()
self.dfa = DFA(dfa_layers=[
self.dfa_atten_embedding, *self.dfa_self_attns, self.dfa_cross,
*self.mlp.dfa_layers, self.dfa_embedding
],
feedback_points_handling=FeedbackPointsHandling.LAST,
no_training=training_method != 'dfa')
class FM(torch.nn.Module):
"""Factorization Machines"""
def __init__(self, opt):
super(FM, self).__init__()
self.use_cuda = opt.get('use_cuda')
self.latent_dim = opt['latent_dim']
self.field_dims = opt['field_dims']
self.feature_num = sum(self.field_dims)
self.embedding = PEPEmbedding(opt)
self.linear = FeaturesLinear(self.field_dims) # linear part
self.fm = FactorizationMachine(reduce_sum=True)
print("BackBone Embedding Parameters: ", self.feature_num * self.latent_dim)
def forward(self, x):
linear_score = self.linear.forward(x)
xv = self.embedding(x)
fm_score = self.fm.forward(xv)
score = linear_score + fm_score
return score.squeeze(1)
def l2_penalty(self, x, lamb):
xv = self.embedding(x)
xv_sq = xv.pow(2)
xv_penalty = xv_sq * lamb
xv_penalty = xv_penalty.sum()
return xv_penalty
def calc_sparsity(self):
base = self.feature_num * self.latent_dim
non_zero_values = torch.nonzero(self.embedding.sparse_v).size(0)
percentage = 1 - (non_zero_values / base)
return percentage, non_zero_values
def get_threshold(self):
return self.embedding.g(self.embedding.s)
def get_embedding(self):
return self.embedding.sparse_v.detach().cpu().numpy()
class LR(torch.nn.Module):
def __init__(self, opt):
super(LR, self).__init__()
self.use_cuda = opt.get('use_cuda')
self.field_dims = opt['field_dims']
self.linear = FeaturesLinear(self.field_dims) # linear part
def forward(self, x):
"""Compute Score"""
score = self.linear.forward(x)
return score.squeeze(1)
def l2_penalty(self, x, lamb):
return 0
def calc_sparsity(self):
return 0, 0
def get_threshold(self):
return 0
def get_embedding(self):
return np.zeros(1)
def __init__(self,
field_dims,
embed_dim,
LNN_dim,
mlp_dims,
dropouts,
training_method='dfa'):
super().__init__()
self.num_fields = len(field_dims)
self.linear = FeaturesLinear(field_dims) # Linear
self.embedding = FeaturesEmbedding(field_dims, embed_dim) # Embedding
self.dfa_embed = DFALayer()
self.LNN_dim = LNN_dim
self.LNN_output_dim = self.LNN_dim * embed_dim
self.LNN = LNN(self.num_fields, embed_dim, LNN_dim)
self.dfa_lnn = DFALayer()
self.mlp = DFAMultiLayerPerceptron(self.LNN_output_dim,
mlp_dims,
dropouts[0],
dfa_output=False)
self.dfa = DFA(dfa_layers=[self.dfa_lnn, *self.mlp.dfa_layers],
feedback_points_handling=FeedbackPointsHandling.LAST,
no_training=training_method != 'dfa')
def __init__(self, field_dims, embed_dim):
super().__init__()
self.linear = FeaturesLinear(field_dims)
self.ffm = FieldAwareFactorizationMachine(field_dims, embed_dim)
def __init__(self, field_dims, t, lam):
super().__init__()
self.linear = FeaturesLinear(field_dims, t, lam)
def __init__(self, opt):
super(LR, self).__init__()
self.use_cuda = opt.get('use_cuda')
self.field_dims = opt['field_dims']
self.linear = FeaturesLinear(self.field_dims) # linear part
