The benefit of machine learning is well understood now, but for a learning system to perform accurately it needs to learn from suitable datasets of information. Often such information are a threat to privacy. For example, a machine learning algorithm to suggest friends in social networking site needs to look at personal data which can be considered a breach of privacy to many people. Although absolute privacy will be preserved if such information are not accessed, that will prevent the development of machine learning algorithms. Similar problems exist in many real world situations.
Differential privacy comes with an elegant provable solution to protect individual's privacy in spite of making the dataset of information available to machine learning algorithms. Differential privacy is an emerging field which has already drawn considerable interest in leading machine learning communities such as NIPS and ICML as well as journals like Science.
However, differential privacy is a relatively new field and considerably challenging to apply with machine learning. There is a huge potential from research perspective. This course intends to provide a gateway to this interesting topic through basic principles as well as state of the art papers.
First, a few lectures will be given to motivate and introduce the topic. After that, each student will present a paper, explaining the connection with the previous presentation, background of the paper, main concept of the paper, and listing all relevant papers, etc. Following each presentation there will be a discussion and quiz session encouraging active participation to the topic.
The course is mainly aimed at doctoral students and researchers but may also be of interest to advanced master's students. Note that due to the format of the course, the number of students is limited to maximum 20.
Mrinal Das, Tomi Peltola
Guest lecturer Marta Soare
Advisor Samuel Kaski
3 for presenting one paper
5 for presenting one paper and reporting results of the paper after implementing the concept
8 for presenting one paper and one mini-project
Basic Mathematics, Machine Learning basic principles