Skip to content

Bayesian Neural Networks

ToDos

  1. Get Baseline Models + Uncertainty
  2. Baseline NNs w. 1D Inputs
  3. Uncertainty

Baselines

This just means using some of the typical sklearn models and learning and doing predictions.


Baseline NNs w. 1D Inputs

We can base our work off of a recent paper (Code). They were able to use a 1D CNN architecture and got some good results. So let's replicate that!


SOTA Models w. DropOut

  • VGG
  • PreResNet
  • Wide ResNet
  • Tiramisu

Stochastic Weight Averaging (SWA)


  • Uncertainty Baselines with edward2 - repo
  • Uncertainty Baselines with TensorFlow (OatML) - repo
  • Understanding BDL with PyTorch - repo
  • Non-Bayesian Benchmarks - repo