Uncertainty Quantification in Deep Learning Methods

Master

This topic focuses on developing methods to estimate the uncertainties in deep learning models during training and inference time. The subject involves applying probabilistic and information theoric methods on both datasets and models to quantify the unknown situations. We can provide access to real systems to apply the proposed methods.

Goal

To discover and uncertainties in the input datasets and models to increase the prediction performance.

Learning outcome

You will learn machine learning/deep learning techniques with uncertainty aspects.

Qualifications

Some programming (especially in Python) and modeling skills. Knowledge of machine learning frameworks such as Tensorflow and Keras is an added advantage.

Supervisors

  • Shaukat Ali
  • Ferhat Ozgur Catak

References

[1] Gal, Yarin, and Zoubin Ghahramani. "Dropout as a bayesian approximation: Representing model uncertainty in deep learning." international conference on machine learning. 2016.
[2] Maddox, W. J., Izmailov, P., Garipov, T., Vetrov, D. P., & Wilson, A. G. (2019). A simple baseline for bayesian uncertainty in deep learning. In Advances in Neural Information Processing Systems (pp. 13153-13164).
[3] Kuleshov, V., Fenner, N., & Ermon, S. (2018). Accurate uncertainties for deep learning using calibrated regression. arXiv preprint arXiv:1807.00263.

Contact person