• Bayesian Deep Learning Reading list
  • Preface
  • Machine Learning
    • Uncertainty Quantization
    • Statistic Models
  • Computer Vision
    • Generative model: AIR
    • Generative models: VAE & GAN
  • Time Series

DBL is a sort of model that integrating statistical property into Deep models to calibrate the confidence corresponding to the data and modles. Depending on the experimental setting, we can divide the DBL models into three parts:

• Generative models in CV
• Uncertainty quantization in Machine Learning
  • partially or completely focus on OoD and Active Learning Tasks
  • Stochastic Process Models and BNNs
• Time series Problems

Many works have been done to analyze the connection between Deep models and Statistic property, including:

• Surveys

  • Uncertainty in Deep Learning
  • A Survey on Bayesian Deep Learning

• Workshops:

  • NeurIPS
  • Gaussian Process Summer Schools

• Scholars & Groups:

  • OATML: Oxford Applied And Theoretical Machine Learning Group

  • Andrew Gordon Wilson, NYU

  • Neil D. Lawrence, Cantab

  • Balaji Lakshminarayanan

Here Thesis marked in $\spadesuit$ are some thesis for OoD and $\clubsuit$ for Activate Learning Tasks, and $\blacklozenge$ are in statistic perspective(based on Bayesain neural networks and gaussian process).

• Arxiv21: Uncertainty Baselines: Benchmarks for uncertainty & robustness in deep learning Paper, Code

• NIPS17: What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision? Paper, Code1, Code2

• ICML20-DUQ: Uncertainty Estimation Using a Single Deep Deterministic Neural Network Paper,Code1 , Code2

• $\spadesuit$ NeurIPS20: Energy-based Out-of-distribution Detection Paper,Code

• Arxiv21-DDU: Deterministic Neural Networks with Appropriate Inductive Biases Capture Epistemic and Aleatoric Uncertainty Paper,Code

• Arxiv21-DUE: Improving Deterministic Uncertainty Estimation in Deep Learning for Classification and Regression Paper, Code

• $\blacklozenge$ AISTATS19: Calibrating Deep Convolutional Gaussian Processes Paper,Code

• $\blacklozenge$ NeurIPS20: Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance Awareness Paper,Code