UAI

We will track and plan our work using a GitHub project: https://github.com/orgs/AI-UWr/projects/1/

Basic:

• Gawlikowski, J., Tassi, C.R.N., Ali, M. et al. A survey of uncertainty in deep neural networks. Artif Intell Rev 56 (Suppl 1), 1513–1589 (2023). https://doi.org/10.1007/s10462-023-10562-9 

• Wang et al., Uncertainty in Graph Neural Networks: A Survey, Transactions on Machine Learning Research, 2024, https://openreview.net/forum?id=0e1Kn76HM1

• Bayesian Neural Notworks

• Monte Carlo Dropout

• Bagging, Boosting

• Model Calibration

• Ensembles

• Ensamble Destillation: https://arxiv.org/abs/1503.02531 

• Test-time augmentations

• Hendrycks et al., A BASELINE FOR DETECTING MISCLASSIFIED AND OUT-OF-DISTRIBUTION EXAMPLES IN NEURAL NETWORKS, ICLR 2017, https://arxiv.org/pdf/1610.02136

• Temperature scaling: https://proceedings.mlr.press/v70/guo17a.html 

• https://arxiv.org/abs/2106.04015 

Selected advanced topics:

• https://openreview.net/forum?id=Hkg4TI9xl

• https://openaccess.thecvf.com/content_CVPR_2019/papers/Hein_Why_ReLU_Networks_Yield_High-Confidence_Predictions_Far_Away_From_the_CVPR_2019_paper.pdf

• https://openaccess.thecvf.com/content_CVPRW_2020/papers/w20/Gustafsson_Evaluating_Scalable_Bayesian_Deep_Learning_Methods_for_Robust_Computer_Vision_CVPRW_2020_paper.pdf

• Classification: https://papers.nips.cc/paper_files/paper/2018/hash/a981f2b708044d6fb4a71a1463242520-Abstract.html

• Prior networks: https://papers.nips.cc/paper_files/paper/2018/hash/3ea2db50e62ceefceaf70a9d9a56a6f4-Abstract.html, https://dl.acm.org/doi/10.5555/3454287.3455590

• Ensambles: https://proceedings.neurips.cc/paper_files/paper/2017/hash/9ef2ed4b7fd2c810847ffa5fa85bce38-Abstract.html, https://openreview.net/pdf/b89ef269141324b94996984453d8c77e2a828d57.pdf

• https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Szegedy_Rethinking_the_Inception_CVPR_2016_paper.pdf 

Frameworks:

• https://pyro.ai/

• https://github.com/scikit-learn-contrib/MAPIE/

• https://github.com/deel-ai/puncc

• https://github.com/TyXe-BDL/TyXe

• https://github.com/IntelLabs/bayesian-torch

• https://github.com/TorchUQ/torchuq

• https://pypi.org/project/UncertainSCI/

• https://unco3892.github.io/UncertaintyPlayground/

• others

1. We will develop develop 3 group projects in parallel during the course. You can expect to participate in tasks related to all of them.

2. The goal is to develop our skills and knowledge while developing the projects, so you can be asked to prepare a theoretical presentation or a hands-on tutorial.

3. During each meeting, specific weekly tasks will be assigned to individual students or small groups. Weekly presentation of your results will be the basis for the final grade. 

4. Each week, 3 students will be assigned to serve as Scribes (as a weekly task). Is scribe a good situation? You will have to write down task specifications for your colleagues, review their solutions, and take care of progress tracking. 

5. Maximum of 3 absences are allowed. If absent, a student must arrange for someone (e.g. the Scribe or team members) to present his/her results to get points for the work done. 

6. Delayed tasks are carried over to the following week, unless significant progress has been made indicating that two tasks can be completed in one week. 

7. You can get 13 points for progress reports and 4 points for (optional) advanced presentation. Thresholds for each grade are as follows:

- 3.0 -- 8 points

- 3.5 -- 10 points 

- 4.0 -- 11.5 points

- 4.5 -- 13,5 points

- 5.0 -- 15 points