Interpretations are useful: penalizing explanations to align neural networks with prior knowledge.
Laura RiegerChandan SinghW. James MurdochBin YuPublished in: CoRR (2019)
Keyphrases
- neural network
- prior knowledge
- artificial neural networks
- fuzzy logic
- pattern recognition
- bayesian framework
- rule extraction
- genetic algorithm
- neural network model
- incorporation of prior knowledge
- multilayer perceptron
- feed forward
- back propagation
- database
- domain knowledge
- fuzzy neural network
- network architecture
- fuzzy systems
- multi layer
- generating explanations
- data sets
- learning rules
- activation function
- graph cuts
- neural nets
- associative memory
- decision making
- image processing
- reinforcement learning
- background knowledge