GeMP-BNN: High-Performance Sampling-Free Bayesian Neural Network Accelerator with Gaussian Error Moment Propagation.
Yuki HirayamaKengo SugaShinya Takamaeda-YamazakiPublished in: MCSoC (2023)
Keyphrases
- neural network
- maximum likelihood
- artificial neural networks
- back propagation
- sequential monte carlo
- neural network model
- fault diagnosis
- taylor series expansion
- random sampling
- error rate
- bayesian networks
- akaike information criterion
- monte carlo
- activation function
- pattern recognition
- genetic algorithm
- compute intensive
- neural network is trained
- sampling algorithm
- markov chain monte carlo
- associative memory
- fuzzy logic
- sample size
- gaussian mixture model
- support vector
- neural nets
- density function
- self organizing maps
- feedforward neural networks
- nonlinear functions
- active learning
- broadcast news
- linear complexity
- prediction model
- radial basis function
- feed forward
- hidden layer
- recurrent neural networks
- gaussian distribution
- bp neural network