Awesome

uncertainty-deep-learning

• Simple baselines for uncertainty estimation and model calibration
  1. Bayes by Backprop(ICML 2015) - Weight Uncertainty in Neural Networks
  2. Dropout as a Bayesian Approximation(ICML 2016) https://arxiv.org/abs/1506.02142
  3. Deep Ensembles(NIPS 2017) Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
  4. Temperature scaling (ICML 2017) On Calibration of Modern Neural Networks
• Dependencies
  • Python 3.6+
  • PyTorch==1.1
  • Codes are heavily inspired by https://joshfeldman.net, medium.com/@albertoarrigoni, github.com/gpleiss/temperature_scaling, www.jessicayung.com

Reference

1. Weight Uncertainty in Neural Networks
2. https://arxiv.org/abs/1506.02142
3. What My Deep Model Doesn't Know, Yarin Gal
4. Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
5. On Calibration of Modern Neural Networks