Detecting Out-of-Distribution Inputs in Deep Neural Networks Using an Early-Layer Output.
Vahdat AbdelzadKrzysztof CzarneckiRick SalayTaylor DenoudenSachin VernekarBuu PhanPublished in: CoRR (2019)
Keyphrases
- neural network
- output layer
- multi layer
- hidden layer
- input pattern
- hidden units
- feed forward
- multiple layers
- multilayer perceptron
- artificial neural networks
- probability distribution
- input variables
- desired output
- single layer
- feed forward neural networks
- fuzzy logic
- input data
- back propagation
- pattern recognition
- extreme values
- feedforward neural networks
- neural network model
- fault diagnosis
- uniformly distributed
- rbf network
- fuzzy inference
- automatic detection
- gaussian distribution
- deep learning
- application layer
- random variables
- multi layer perceptron
- maximum likelihood
- spatial distribution