Public defence in Computer Science, M.Sc. Martino Ciaperoni

Title of the doctoral thesis: Efficient and trustworthy methods for knowledge discovery

Doctoral thesis: Martino Ciaperoni
Opponent: Prof. Angela Bonifati, Lyon University, Ranska
Custos: AdjunctProf. Aristides Gionis, Aalto University School of Science, Department of Computer Science

Today, massive amounts of data are collected from a variety of sources and processed at dramatic speed. In this landscape, the field of knowledge discovery, which aims at extracting valuable and actionable knowledge from data, plays a pivotal role in driving decision-making and fostering advancements across various domains. 

Recent advances in knowledge discovery focus on obscure algorithmic methods that are efficient enough to handle the complexity and volume of modern datasets, but are difficult for users to trust. This inability to build trust in obscure algorithmic methods represents a crucial barrier for the adoption of data-driven approaches to decision making, particularly in high-stakes settings. On the other hand, many more traditional approaches to knowledge discovery are more transparent and easily trusted by users, but they often struggle with the challenges posed by modern datasets. 

The doctoral thesis titled "Efficient and trustworthy methods for knowledge discovery" aims at demonstrating that, leveraging a combination of probabilistic and data-management methods, it is possible to design novel algorithmic approaches that are efficient enough to navigate the complexities of the modern world and yet they are easily trusted by users. Thus, the algorithmic methods presented in the thesis are particularly suited for applications in domains where trusting algorithms is an important requirement, such as in high-stakes settings. 

The thesis focuses on four knowledge-discovery tasks that arise naturally in today’s world, namely inference in Bayesian networks, inference in hidden Markov models, multi-label classification and community search in time-varying networks. Nevertheless, the fundamental principles underlying the algorithmic methods proposed in the thesis are general and may be extended to different tasks, contributing to shape the future of knowledge discovery and data-driven algorithmic decision-making.

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact details:

Doctoral theses in the School of Science: https://aaltodoc.aalto.fi/handle/123456789/52

Key words: Knowledge discovery, probabilistic methods, database management, efficient algorithms, trustworthy algorithms