liveBook · Manning

AnoGAN

As explained in Outlier Detection in Python

AnoGAN is a GAN-based anomaly-detection method that estimates a test observation's latent representation and combines reconstruction discrepancy with discriminator output.

AnoGAN detects anomalies by fitting a GAN, estimating each test point’s latent representation, reconstructing it, and combining reconstruction error with discriminator output. section 16.2.3

Its detection procedure is:

  • Fit a generator and discriminator on normal data. section 16.2.3
  • Use gradient descent separately for every test point to find its latent-space location. section 16.2.3
  • Generate a reconstruction and combine its discrepancy with the discriminator’s prediction to estimate outlierness. section 16.2.3

AnoGAN assumes normal observations have good latent-space representations. Its per-point latent search is substantially slower than BiGAN-style inverse mapping and may be impractical during repeated tuning. section 16.2.3

How Outlier Detection in Python covers it

How deep it goes
Explains GAN fitting, per-test-point latent optimization, reconstruction scoring, and computational limitations.
Examples
Uses normal training data and arbitrary test points reconstructed through the learned generator.
What it assumes
Assumes familiarity with GANs, latent representations, generators, discriminators, and reconstruction error.
What you can do after
Diagnose outliers using reconstruction discrepancies and discriminator outputs, while recognizing prohibitive runtime risks.
How it connects
Connects GANs, latent space, generator, discriminator, reconstruction error, BiGAN, and ALAD.
Cover of Outlier Detection in Python
liveBookmanning.com
Read Outlier Detection in Python on liveBook Get the book at manning.com

This explainer was generated by Manning's knowledge graph from the text of Outlier Detection in Python. The section links point to the parts of the book it draws on.