Guiding Neural Network Initialization via Marginal Likelihood Maximization
This work provides an incremental improvement for neural network initialization, potentially benefiting practitioners by simplifying hyperparameter selection.
This paper proposes a data-driven method for neural network initialization by leveraging the connection between neural networks and Gaussian processes. The method, based on marginal likelihood maximization, achieves near-optimal prediction performance on MNIST classification.
We propose a simple, data-driven approach to help guide hyperparameter selection for neural network initialization. We leverage the relationship between neural network and Gaussian process models having corresponding activation and covariance functions to infer the hyperparameter values desirable for model initialization. Our experiment shows that marginal likelihood maximization provides recommendations that yield near-optimal prediction performance on MNIST classification task under experiment constraints. Furthermore, our empirical results indicate consistency in the proposed technique, suggesting that computation cost for the procedure could be significantly reduced with smaller training sets.