Bayesian Nets

• Deep Prior
• Weight Uncertainty in Neural Networks, e.g. Bayes by Backprop
  • Follow-on paper: Bayesian Hypernetworks
    • Train a neural net to output the distribution of the desired neural net.
    • Once trained, input random noise to sample the parameters of the desired net.
    • A lot of the algorithm has to do with being able to scale up to lots of parameters.
• Variational Dropout and the Local Reparameterization Trick
• Dropout as a bayesian approximation: Representing model uncertainty in deep learning
• An Approximate Bayesian Long Short-Term Memory Algorithm for Outlier Detection
• Probabilistic supervised learning
• Using Deep Neural Network Approximate Bayesian Network
• Bayesian Neural Networks
• Efficient Exploration through Bayesian Deep Q-Networks
  • The last layer is a Bayesian Linear Regression Model
  • Performs better than Double DQN and is easier to implement than others
• Bayesian Uncertainty Estimation for Batch Normalized Deep Networks
  • Similar to the ‘Dropout as a bayesian approximation…’ paper, but using Batch Norm layers to calculate uncertainty
• Variational Inference for Policy Gradient
• Modern Computational Methods for Bayesian Inference — A Reading List