Coupling-based Invertible Neural Networks Are Universal Diffeomorphism Approximators.
Takeshi TeshimaIsao IshikawaKoichi TojoKenta OonoMasahiro IkedaMasashi SugiyamaPublished in: NeurIPS (2020)
Keyphrases
- network architecture
- neural network
- neural network model
- activation function
- artificial neural networks
- back propagation
- pattern recognition
- fault diagnosis
- evolutionary algorithm
- training algorithm
- fuzzy systems
- fuzzy logic
- genetic algorithm
- real time
- reinforcement learning
- image processing
- feed forward
- search engine
- artificial intelligence
- information retrieval
- competitive learning