On the Power and Limitations of Random Features for Understanding Neural Networks.
Gilad YehudaiOhad ShamirPublished in: NeurIPS (2019)
Keyphrases
- neural network
- pattern recognition
- feature extraction
- feature vectors
- classification method
- fault diagnosis
- feature set
- classification accuracy
- relative strengths and weaknesses
- invariant features
- keypoints
- low level
- prior knowledge
- artificial neural networks
- feature space
- object recognition
- similarity measure
- image processing
- genetic algorithm