Interpretable Adversarial Perturbation in Input Embedding Space for Text.
Motoki SatoJun SuzukiHiroyuki ShindoYuji MatsumotoPublished in: CoRR (2018)
Keyphrases
- embedding space
- euclidean space
- manifold learning
- low dimensional
- keywords
- text mining
- dimensionality reduction
- visual information
- input space
- geometric structure
- high dimensional
- data points
- image classification
- learning algorithm
- input data
- high dimensional data
- discriminant analysis
- pattern recognition
- image sequences