claCase / Heteroskedasticity-DNN

Full Log-Likelihood Heteroskedastic Regression with Deep Neural Networks and Tensorflow

Heteroskedasticity Regression

Full Log-Likelihood Heteroskedastic Regression with Deep Neural Networks and Tensorflow
Install required libraries:

python -m pip install -r requirements.txt

Linear

Deterministic parameters with no epistemic uncertainty

python main.py --epochs 50 --droprate 0.2 --reduction 7 --mean "linear"

Stochastic parameters with epistemic uncertainty

python main.py --epochs 50 --epistemic --droprate 0.2 --reduction 7 --mean "linear"

Non Linear

Deterministic parameters with no epistemic uncertainty

python main.py  --epochs 50 --droprate 0.2 --reduction 7 --mean "cosine"

Stochastic parameters with epistemic uncertainty

python main.py --epochs 50 --epistemic --droprate 0.2 --reduction 7 --mean "cosine"