Training Normalizing Flows with the Information Bottleneck for Competitive Generative Classification.
Lynton ArdizzoneRadek MackowiakCarsten RotherUllrich KöthePublished in: NeurIPS (2020)
Keyphrases
- training phase
- supervised learning
- training set
- training process
- discriminative training
- pattern recognition
- training samples
- automatic classification
- pattern classification
- support vector machine
- image classification
- classification rules
- machine learning
- classification systems
- multi layer perceptron
- discriminative learning
- discriminative classifiers
- classification performances
- feature space
- machine learning algorithms
- training algorithm
- text classification
- classification accuracy
- feature vectors
- classification method
- decision trees
- artificial neural networks
- hybrid generative discriminative
- sufficient training data
- avoid overfitting
- conditional likelihood
- neural network
- labelled data
- training speed
- unsupervised learning
- standard deviation
- decision rules
- classification algorithm