1. Feature selection
1.1 Unsupervised Learning
- stacked autoencoders and keep the last hidden layer as your features
- stacked RBMS (Deep Belief Network) and keep the last layer as features
- sparse filtering in several layers
- convolutional deep networks (for images)
References:
Representation Learning: A Review and NewPerspectives, Yoshua Bengio, etc, 2014
Tutorial on variational autoencoders, Carl Doersch, 2016
Feature Extraction: Foundations and applications, Isabelle Guyon, 2006
2. Application topics
2.1 Credit card fraud detection
2.2 Customer Churn prediction
2.3