Various estimators of the infinite dimensional exponential family model. In particular, effecient approximations from our NIPS 2015 paper on Gradient-free Hamiltonain Monte Carlo with Effecient Kernel Exponential Families. Used in the accompanying kernel HMC package.
For learning parameters, there is the option to use the Bayesian optimisation package pybo. If theano is installed, higher order derivatives of the model's log-density are available.
Install dependencies:
pip install -r https://raw.githubusercontent.com/karlnapf/kernel_exp_family/master/requirements.txt
Install kernel_exp_family:
pip install git+https://github.com/karlnapf/kernel_exp_family.git
A list of examples can be found here. For example, run
python -m kernel_exp_family.examples.demo_simple.py