Horseshoe Priors for Bayesian Neural Networks.

---

This software includes an implementation of the model presented in:
"Model Selection in Bayesian Neural Networks via Horseshoe priors", Soumya Ghosh and Finale Doshi-Velez. NIPS 2017. Bayesian Deep Learning Workshop (http://bayesiandeeplearning.org/2017/papers/42.pdf).

The software depends on autograd (https://github.com/HIPS/autograd)

USAGE:

• To train,

python ./scripts/train_hsbnn.py

• You can use

python ./src/plot_weights.py path/to/results.pkl

to visualize the inferred weights.