Department of Computer Science

Summer internship positions in the PROBIC group in 2023

1. Differentially private machine learning (multiple positions)

Supervisor: Antti Honkela

Background: machine learning, mathematics, programming skills

Differentially private machine learning studies learning methods that can operate while guaranteeing privacy of the data subjects. These methods can be applied to solving predictive learning problems on private data but also for creating provably anonymised data sets. We apply differential privacy in the context of various modern machine learning methods, including Bayesian methods, deep learning and federated learning.

In this project you will participate in developing and applying new differentially private machine learning methods. Depending on your background and interests, the work will combine working on the mathematical theory of differential privacy, general methods development, implementation and application of the developed methods in different applications. The work will involve collaboration with Finnish Center for Artificial Intelligence FCAI Research Programme in Privacy-preserving and Secure AI.

The topic is suitable for a Master's thesis topic.

Examples of related MSc theses done in the group:

Marlon Tobaben. Hyperparameters and neural architectures in differentially private deep learning. MSc Thesis, University of Helsinki, 2022.

Ossi Räisä. Differentially Private Metropolis–Hastings Algorithms. MSc Thesis, University of Helsinki, 2021.

Examples of our recent related papers:

Joonas Jälkö, Eemil Lagerspetz, Jari Haukka, Sasu Tarkoma, Antti Honkela, and Samuel Kaski. Privacy-preserving data sharing via probabilistic modeling. Patterns 2(7):100271 (2021).

Tejas Kulkarni, Joonas Jälkö, Antti Koskela, Samuel Kaski, and Antti Honkela. Differentially Private Bayesian Inference for Generalized Linear Models. In Proceedings of the 38th International Conference on Machine Learning (ICML 2021) (2021).

Antti Koskela, Joonas Jälkö, Lukas Prediger, and Antti Honkela. Tight Differential Privacy for Discrete-Valued Mechanisms and for the Subsampled Gaussian Mechanism Using FFT. In Proceedings of the 24th International Conference on Artificial Intelligence and Statistics (AISTATS 2021) (2021).