Source code for pygod.detector.gaan

# -*- coding: utf-8 -*-
"""Generative Adversarial Attributed Network Anomaly Detection (GAAN)"""
# Author: Ruitong Zhang <rtzhang@buaa.edu.cn>, Kay Liu <zliu234@uic.edu>
# License: BSD 2 clause

import torch
import warnings
import torch.nn.functional as F
from torch_geometric.nn import MLP
from torch_geometric.utils import to_dense_adj

from ..nn import GAANBase
from . import DeepDetector


[docs] class GAAN(DeepDetector): """ Generative Adversarial Attributed Network Anomaly Detection GAAN is a generative adversarial attribute network anomaly detection framework, including a generator module, an encoder module, a discriminator module, and uses anomaly evaluation measures that consider sample reconstruction error and real sample recognition confidence to make predictions. This model is transductive only. See :cite:`chen2020generative` for details. Parameters ---------- noise_dim : int, optional Input dimension of the Gaussian random noise. Defaults: ``16``. hid_dim : int, optional Hidden dimension of model. Default: ``64``. num_layers : int, optional Total number of layers in model. A half (floor) of the layers are for the generator, the other half (ceil) of the layers are for encoder. Default: ``4``. dropout : float, optional Dropout rate. Default: ``0.``. weight_decay : float, optional Weight decay (L2 penalty). Default: ``0.``. act : callable activation function or None, optional Activation function if not None. Default: ``torch.nn.functional.relu``. backbone : torch.nn.Module The backbone of GAAN is fixed to be MLP. Changing of this parameter will not affect the model. Default: ``None``. contamination : float, optional The amount of contamination of the dataset in (0., 0.5], i.e., the proportion of outliers in the dataset. Used when fitting to define the threshold on the decision function. Default: ``0.1``. lr : float, optional Learning rate. Default: ``0.004``. epoch : int, optional Maximum number of training epoch. Default: ``100``. gpu : int GPU Index, -1 for using CPU. Default: ``-1``. batch_size : int, optional Minibatch size, 0 for full batch training. Default: ``0``. num_neigh : int, optional Number of neighbors in sampling, -1 for all neighbors. Default: ``-1``. weight : float, optional Weight between reconstruction of node feature and structure. Default: ``0.5``. verbose : int, optional Verbosity mode. Range in [0, 3]. Larger value for printing out more log information. Default: ``0``. save_emb : bool, optional Whether to save the embedding. Default: ``False``. compile_model : bool, optional Whether to compile the model with ``torch_geometric.compile``. Default: ``False``. **kwargs Other parameters for the backbone. Attributes ---------- decision_score_ : torch.Tensor The outlier scores of the training data. Outliers tend to have higher scores. This value is available once the detector is fitted. threshold_ : float The threshold is based on ``contamination``. It is the :math:`N \\times` ``contamination`` most abnormal samples in ``decision_score_``. The threshold is calculated for generating binary outlier labels. label_ : torch.Tensor The binary labels of the training data. 0 stands for inliers and 1 for outliers. It is generated by applying ``threshold_`` on ``decision_score_``. emb : torch.Tensor or tuple of torch.Tensor or None The learned node hidden embeddings of shape :math:`N \\times` ``hid_dim``. Only available when ``save_emb`` is ``True``. When the detector has not been fitted, ``emb`` is ``None``. When the detector has multiple embeddings, ``emb`` is a tuple of torch.Tensor. """ def __init__(self, noise_dim=16, hid_dim=64, num_layers=4, dropout=0., weight_decay=0., act=F.relu, backbone=None, contamination=0.1, lr=4e-3, epoch=100, gpu=-1, batch_size=0, num_neigh=-1, weight=0.5, verbose=0, save_emb=False, compile_model=False, **kwargs): self.noise_dim = noise_dim self.weight = weight # self.num_layers is 1 for sample one hop neighbors # In GAAN, self.model_layers is for model layers self.model_layers = num_layers if backbone is not None: warnings.warn('GAAN can only use MLP as the backbone.') super(GAAN, self).__init__( hid_dim=hid_dim, num_layers=1, dropout=dropout, weight_decay=weight_decay, act=act, contamination=contamination, lr=lr, epoch=epoch, gpu=gpu, batch_size=batch_size, num_neigh=num_neigh, verbose=verbose, gan=True, save_emb=save_emb, compile_model=compile_model, **kwargs) def process_graph(self, data): GAANBase.process_graph(data) def init_model(self, **kwargs): if self.save_emb: self.emb = torch.zeros(self.num_nodes, self.hid_dim) return GAANBase(in_dim=self.in_dim, noise_dim=self.noise_dim, hid_dim=self.hid_dim, num_layers=self.model_layers, dropout=self.dropout, act=self.act, **kwargs).to(self.device) def forward_model(self, data): batch_size = data.batch_size node_idx = data.n_id x = data.x.to(self.device) s = data.s.to(self.device) edge_index = data.edge_index.to(self.device) noise = torch.randn(x.shape[0], self.noise_dim).to(self.device) x_, a, a_ = self.model(x, noise) loss_g = self.model.loss_func_g(a_[edge_index]) self.opt_in.zero_grad() loss_g.backward() self.opt_in.step() self.epoch_loss_in += loss_g.item() * batch_size loss = self.model.loss_func_ed(a[edge_index], a_[edge_index].detach()) score = self.model.score_func(x=x[:batch_size], x_=x_[:batch_size], s=s[:batch_size, node_idx], s_=a[:batch_size], weight=self.weight, pos_weight_s=1, bce_s=True) return loss, score.detach().cpu()