Negative eigenvalues of the Hessian in deep neural networks.
Guillaume AlainNicolas Le RouxPierre-Antoine ManzagolPublished in: ICLR (Workshop) (2018)
Keyphrases
- neural network
- pattern recognition
- covariance matrix
- artificial neural networks
- fuzzy logic
- genetic algorithm
- back propagation
- positive and negative
- neural network model
- artificial intelligence
- feed forward
- covariance matrices
- neural nets
- recurrent neural networks
- fault diagnosis
- principal components
- principal component analysis
- associative memory
- feature extraction
- fuzzy systems
- training process
- learning rules
- learning algorithm
- rule extraction
- feedforward neural networks
- deep learning
- eigendecomposition
- adaptive resonance theory
- neural network is trained