Simple numpy implementation of (Bayesian) Neural Network
Running MNIST experiment
- change the param.path_data in demo/mnist-sghmc.py to your dataset
- python mnist-sghmc.py 2>log.txt
- note: the evaluation statistics comes out from stderr, which is redirected into log.txt in last command
How to adapting to other classification experiment:
- check mnist.py, which is specific to mnist dataset experiment
- get a copy, change the function cfg_param and run_exp
- basically we need a rewritten version of mnist.py and demo/mnist-sghmc.py