Classified as unknown: A novel Bayesian neural network
This work provides an incremental improvement for researchers and practitioners needing efficient Bayesian neural networks, generalizing a prior binary classification method to multi-class settings.
The authors tackled the problem of computationally expensive Bayesian neural network training by developing a closed-form Bayesian learning algorithm that eliminates gradient calculations and Monte Carlo sampling, achieving sequential learning for multi-class classification with multi-layer perceptrons.
We establish estimations for the parameters of the output distribution for the softmax activation function using the probit function. As an application, we develop a new efficient Bayesian learning algorithm for fully connected neural networks, where training and predictions are performed within the Bayesian inference framework in closed-form. This approach allows sequential learning and requires no computationally expensive gradient calculation and Monte Carlo sampling. Our work generalizes the Bayesian algorithm for a single perceptron for binary classification in \cite{H} to multi-layer perceptrons for multi-class classification.