forked from JointCloud/JCC-DeepOD
275 lines
9.7 KiB
Python
275 lines
9.7 KiB
Python
# -*- coding: utf-8 -*-
|
|
"""
|
|
Deep anomaly detection with deviation networks.
|
|
PyTorch's implementation
|
|
@Author: Hongzuo Xu <hongzuoxu@126.com, xuhongzuo13@nudt.edu.cn>
|
|
"""
|
|
|
|
from deepod.core.base_model import BaseDeepAD
|
|
from deepod.core.networks.base_networks import get_network
|
|
from deepod.models.tabular.devnet import DevLoss
|
|
from torch.utils.data import DataLoader, TensorDataset
|
|
from torch.utils.data.sampler import WeightedRandomSampler
|
|
import torch
|
|
import numpy as np
|
|
|
|
|
|
class DevNetTS(BaseDeepAD):
|
|
"""
|
|
Deviation Networks for Weakly-supervised Anomaly Detection (KDD'19)
|
|
:cite:`pang2019deep`
|
|
|
|
Deviation Networks (DevNet) designed for weakly-supervised anomaly detection.
|
|
This implementation is based on the architecture presented in the KDD'19 paper:
|
|
"Deviation Networks for Weakly-supervised Anomaly Detection" by Pang et al.
|
|
|
|
Args:
|
|
|
|
hidden_dims (Union[list, str, int], optional):
|
|
The dimensions for the hidden layers. Can be a list of integers, a string of comma-separated integers, or a single integer.
|
|
- If list, each item is a layer
|
|
- If str, neural units of hidden layers are split by comma
|
|
- If int, number of neural units of single hidden layer
|
|
- Defaults to '100,50'.
|
|
|
|
act (str, optional):
|
|
Activation function to use. Choices include 'ReLU', 'LeakyReLU', 'Sigmoid', 'Tanh'. Default is 'ReLU'.
|
|
|
|
bias (bool, optional):
|
|
Whether to include a bias term in the linear layers. Default is False.
|
|
|
|
n_heads (int, optional):
|
|
Number of heads in multi-head attention. Only used when network is 'transformer'. Default is 8.
|
|
|
|
pos_encoding (str, optional):
|
|
The type of positional encoding to use. Only relevant when network is 'transformer'. Choices are 'fixed' or 'learnable'. Default is 'fixed'.
|
|
|
|
norm (str, optional):
|
|
Normalization method in the Transformer. Only relevant when network is 'transformer'. Choices are 'LayerNorm' or 'BatchNorm'. Default is 'LayerNorm'.
|
|
|
|
epochs (int, optional):
|
|
Number of training epochs. Default is 100.
|
|
|
|
batch_size (int, optional):
|
|
Batch size for training. Default is 64.
|
|
|
|
lr (float, optional):
|
|
Learning rate for the optimizer. Default is 1e-3.
|
|
|
|
network (str, optional):
|
|
Type of network architecture to use. Default is 'Transformer'.
|
|
|
|
seq_len (int, optional):
|
|
Length of input sequences for models that require it. Default is 100.
|
|
|
|
stride (int, optional):
|
|
Stride of the convolutional layers. Default is 1.
|
|
|
|
rep_dim (int, optional):
|
|
The representation dimension. Unused in this model but kept for consistency. Default is 128.
|
|
|
|
d_model (int, optional):
|
|
The number of expected features in the transformer model. Only used when network is 'transformer'. Default is 512.
|
|
|
|
attn (str, optional):
|
|
Type of attention to use. Only used when network is 'transformer'. Default is 'self_attn'.
|
|
|
|
margin (float, optional):
|
|
Margin for the deviation loss function. Default is 5.
|
|
|
|
l (int, optional):
|
|
The size of the sample for the Gaussian distribution in the deviation loss function. Default is 5000.
|
|
|
|
epoch_steps (int, optional):
|
|
Maximum number of steps per epoch. If -1, all batches will be processed. Default is -1.
|
|
|
|
prt_steps (int, optional):
|
|
Number of epoch intervals for printing during training. Default is 10.
|
|
|
|
device (str, optional):
|
|
The device to use for training ('cuda' or 'cpu'). Default is 'cuda'.
|
|
|
|
verbose (int, optional):
|
|
Verbosity mode. 0 = silent, 1 = progress bar, 2 = one line per epoch. Default is 2.
|
|
|
|
random_state (int, optional):
|
|
Seed for the random number generator for reproducibility. Default is 42.
|
|
|
|
"""
|
|
|
|
def __init__(self, epochs=100, batch_size=64, lr=1e-3,
|
|
network='Transformer', seq_len=100, stride=1,
|
|
rep_dim=128, hidden_dims='100,50', act='ReLU', bias=False,
|
|
n_heads=8, d_model=512, attn='self_attn', pos_encoding='fixed', norm='LayerNorm',
|
|
margin=5., l=5000,
|
|
epoch_steps=-1, prt_steps=10, device='cuda',
|
|
verbose=2, random_state=42):
|
|
"""
|
|
Initialize the DevNetTS.
|
|
"""
|
|
super(DevNetTS, self).__init__(
|
|
data_type='ts', model_name='DevNet', epochs=epochs, batch_size=batch_size, lr=lr,
|
|
network=network, seq_len=seq_len, stride=stride,
|
|
epoch_steps=epoch_steps, prt_steps=prt_steps, device=device,
|
|
verbose=verbose, random_state=random_state
|
|
)
|
|
|
|
self.margin = margin
|
|
self.l = l
|
|
|
|
self.hidden_dims = hidden_dims
|
|
self.act = act
|
|
self.bias = bias
|
|
|
|
# parameters for Transformer
|
|
self.n_heads = n_heads
|
|
self.d_model = d_model
|
|
self.attn = attn
|
|
self.pos_encoding = pos_encoding
|
|
self.norm = norm
|
|
|
|
return
|
|
|
|
def training_prepare(self, X, y):
|
|
"""
|
|
Prepares the data and model for training by creating a balanced data loader,
|
|
initializing the network, and setting up the loss criterion.
|
|
|
|
Args:
|
|
|
|
X (np.ndarray):
|
|
The input features for training.
|
|
|
|
y (np.ndarray):
|
|
The target labels for training, where 1 indicates an anomaly.
|
|
|
|
Returns:
|
|
|
|
train_loader (DataLoader):
|
|
A DataLoader with balanced mini-batches for training.
|
|
|
|
net (nn.Module):
|
|
The initialized neural network model.
|
|
|
|
criterion (Loss):
|
|
The loss function used during training.
|
|
|
|
"""
|
|
|
|
# loader: balanced loader, a mini-batch contains a half of normal data and a half of anomalies
|
|
n_anom = np.where(y == 1)[0].shape[0]
|
|
n_norm = self.n_samples - n_anom
|
|
weight_map = {0: 1. / n_norm, 1: 1. / n_anom}
|
|
|
|
dataset = TensorDataset(torch.from_numpy(X).float(), torch.from_numpy(y).long())
|
|
sampler = WeightedRandomSampler(weights=[weight_map[label.item()] for data, label in dataset],
|
|
num_samples=len(dataset), replacement=True)
|
|
train_loader = DataLoader(dataset, batch_size=self.batch_size, sampler=sampler)
|
|
|
|
network_params = {
|
|
'n_features': self.n_features,
|
|
'n_hidden': self.hidden_dims,
|
|
'n_output': 1,
|
|
'activation': self.act,
|
|
'bias': self.bias
|
|
}
|
|
if self.network == 'Transformer':
|
|
network_params['n_heads'] = self.n_heads
|
|
network_params['d_model'] = self.d_model
|
|
network_params['pos_encoding'] = self.pos_encoding
|
|
network_params['norm'] = self.norm
|
|
network_params['attn'] = self.attn
|
|
network_params['seq_len'] = self.seq_len
|
|
elif self.network == 'ConvSeq':
|
|
network_params['seq_len'] = self.seq_len
|
|
|
|
network_class = get_network(self.network)
|
|
net = network_class(**network_params).to(self.device)
|
|
|
|
criterion = DevLoss(margin=self.margin, l=self.l)
|
|
|
|
if self.verbose >= 2:
|
|
print(net)
|
|
|
|
return train_loader, net, criterion
|
|
|
|
def inference_prepare(self, X):
|
|
"""
|
|
Prepares the data for inference.
|
|
|
|
Args:
|
|
|
|
X (Tensor):
|
|
The input features for inference.
|
|
|
|
Returns:
|
|
|
|
test_loader (DataLoader):
|
|
A DataLoader for inference.
|
|
|
|
"""
|
|
|
|
test_loader = DataLoader(X, batch_size=self.batch_size,
|
|
drop_last=False, shuffle=False)
|
|
self.criterion.reduction = 'none'
|
|
return test_loader
|
|
|
|
def training_forward(self, batch_x, net, criterion):
|
|
"""
|
|
Performs a forward pass during training.
|
|
|
|
Args:
|
|
|
|
batch_x (tuple):
|
|
A batch of input features and target labels.
|
|
|
|
net (nn.Module):
|
|
The neural network model.
|
|
|
|
criterion (Loss):
|
|
The loss function used during training.
|
|
|
|
Returns:
|
|
|
|
loss (Tensor):
|
|
The computed loss for the batch.
|
|
"""
|
|
|
|
batch_x, batch_y = batch_x
|
|
batch_x = batch_x.float().to(self.device)
|
|
batch_y = batch_y.to(self.device)
|
|
pred = net(batch_x)
|
|
loss = criterion(batch_y, pred)
|
|
return loss
|
|
|
|
def inference_forward(self, batch_x, net, criterion):
|
|
"""
|
|
Performs a forward pass during inference.
|
|
|
|
Args:
|
|
|
|
batch_x (Tensor):
|
|
A batch of input features.
|
|
|
|
net (nn.Module):
|
|
The neural network model.
|
|
|
|
criterion (Loss):
|
|
The loss function used during training. Not used.
|
|
|
|
Returns:
|
|
|
|
batch_z (Tensor):
|
|
The batch of input features (unmodified).
|
|
|
|
s (Tensor):
|
|
The computed scores for the batch.
|
|
|
|
"""
|
|
|
|
batch_x = batch_x.float().to(self.device)
|
|
s = net(batch_x)
|
|
s = s.view(-1)
|
|
batch_z = batch_x
|
|
return batch_z, s
|