autoqild.classifiers

This package contains Blind classifiers, which just uses the output class values to make predictions by learning a constant function, used information leakage detection.

We also implement bayes predictor which uses the orignal PDF generating the datasets to get the predictions. Additonally, Multi Layer Perceptron is implemented using keras.

Modules

bayes_predictor	A Bayes-optimal classifier leveraging the underlying probability distribution of the dataset for predictions.
blind_classifiers	Classes implementing classifiers which predicts a constant function which predict label only using the outputs of the dataset, which are used as baselines for detecting information leakage.
layers	A custom dense layer with batch normalization and optional pre-activation normalization.
multi_layer_perceptron	A customizable feedforward neural network implemented using Keras for more complex classification tasks.