liveBook · Manning

AEGMM

As explained in Outlier Detection in Python

AEGMM is a model that jointly trains an autoencoder and a Gaussian mixture model with a shared objective for representation learning and outlier detection.

AEGMM jointly trains an autoencoder and a Gaussian mixture model to learn representations and detect outliers. The autoencoder compresses inputs and measures reconstruction error; the GMM evaluates the resulting embedding. section 16.4

Its detection logic combines two signals:

  • The autoencoder maps inputs into a low-dimensional representation and reconstructs them. section 16.4
  • The GMM evaluates the learned embedding for outlier status. section 16.4
  • Joint training uses a shared objective instead of fitting the autoencoder and GMM separately. section 16.4

Noisy CIFAR10 image

A CIFAR10 image is reconstructed closely when unaltered. After random pixel noise is added, reconstruction error rises and the image is flagged as an outlier. 

pip install alibi-detect

Install alibi-detect before configuring deep outlier detectors. section 16.4

How Outlier Detection in Python covers it

How deep it goes
Explains the representation, reconstruction, GMM, and shared objective, but provides no AEGMM-specific implementation.
Examples
Uses CIFAR10 images, close reconstructions, and randomly pixel-perturbed images.
What it assumes
Assumes familiarity with autoencoders, GMM expectation maximization, and TensorFlow or PyTorch installation.
What you can do after
Apply it via alibi-detect after installing TensorFlow or PyTorch and configuring a deep outlier detector.
How it connects
Connects autoencoders, reconstruction error, Gaussian mixture models, VAEGMM, and outlier detection.
Cover of Outlier Detection in Python
liveBookmanning.com
Read Outlier Detection in Python on liveBook Get the book at manning.com

Related concepts

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.