Anomaly Detection with Autoencoders: Better Than Rules, Cheaper Than Labels

Why Autoencoders for Anomaly Detection

You only need normal data to train. The autoencoder learns to reconstruct normal patterns well. Anomalies have high reconstruction error.

Architecture

class Autoencoder(nn.Module):
    def __init__(self, input_dim, latent_dim=32):
        super().__init__()
        self.encoder = nn.Sequential(
            nn.Linear(input_dim, 128), nn.ReLU(),
            nn.Linear(128, 64), nn.ReLU(),
            nn.Linear(64, latent_dim)
        )
        self.decoder = nn.Sequential(
            nn.Linear(latent_dim, 64), nn.ReLU(),
            nn.Linear(64, 128), nn.ReLU(),
            nn.Linear(128, input_dim)
        )
    
    def forward(self, x):
        z = self.encoder(x)
        return self.decoder(z)

Threshold Setting

# Set threshold at 99th percentile of train reconstruction error
recon_errors = ((X_train - model(X_train)) ** 2).mean(dim=1)
threshold = np.percentile(recon_errors.numpy(), 99)