This is an informal implementation of [1711.00165] Deep Neural Networks as Gaussian Processes.

python3 main.py -mode ["dnn" or "rbf"] -iter_size [int]

for example

python3 main.py -mode dnn -iter_size 20

To optimize hyper parameters of kernels, We use MCMC sampling. In this code, We adopted Metropolis-Haistings.