Opening the Black Box: predicting the trainability of deep neural networks with reconstruction entropy.
Yanick ThurnRo JeffersonJohanna ErdmengerPublished in: CoRR (2024)
Keyphrases
- black box
- neural network
- rule extraction
- hybrid systems
- black boxes
- white box
- three dimensional
- artificial neural networks
- integration testing
- pattern recognition
- information theory
- information theoretic
- back propagation
- image reconstruction
- neural network model
- recurrent neural networks
- fuzzy logic
- test cases
- state transition
- white box testing
- data sets
- multilayer perceptron
- mutual information
- feedforward neural networks
- multimedia
- metadata
- genetic algorithm