Keynotes and ICDT Invited Talk

Keynote 1

Title: The Quest for Knowledge

Katja Hose

Aalborg University, Denmark

Abstract: 

Throughout the entire history of mankind, humans have always strived to acquire new knowledge. In a way, striving for knowledge is still what unites researchers across all modern research disciplines. Yet, every single one of them has a slightly different understanding. Read More

Keynote 2

Title: Explainability Queries for ML Models and its Connections with Data Management Problems

Pablo Barceló

Pontifical Catholic University of Chile, Chile

Abstract:

In this talk I will present two recent examples of my research on explainability problems over machine learning (ML) models. In rough terms, these explainability problems deal with specific queries one poses over a ML model in order to obtain meaningful justifications for their results. Read More

Keynote 3

Title: Data Profiling – A look back and a look forward

Felix Naumann

Hasso Plattner Institute (HPI), Germany

Abstract:

Data profiling is the act of extracting many different types of metadata from a given dataset. This research area has recently thrived, due to (i) its simple problem statements, such as “discover all key candidates”, (ii) the high computational complexity of the problems, which are often exponential in the number of columns, (iii) the manifold opportunities for optimizations, such as apriori-inspired pruning or data sampling, and (iv) the various application areas for data profiling results, such as query optimization and data cleaning. Read More

Keynote  4

Title: Comparing Apples and Oranges: Fairness and Diversity in Ranking

Julia Stoyanovich

New York University, USA

Abstract:

Algorithmic rankers take a collection of candidates as input and produce a ranking (permutation) of the candidates as output. The simplest kind of ranker is score-based; it computes a score of each candidate independently and returns the candidates in score order.  Another common kind of ranker is learning-to-rank, where supervised learning is used to predict the ranking of unseen candidates.  For both kinds of rankers, we may output the entire permutation or only the highest scoring k candidates, the top-k.  Set selection is a special case  of ranking that ignores the relative order among the top-k.Read More

ICDT Invited Talk

Title: What Makes a Variant of Query Determinacy (Un) Decidable?

Jerzy Marcinkowski

University of Wrocław, Poland

Abstract:

Suppose there is a database we have no direct access to, but there are views of this database available to us, defined by some queries Q_1, Q_2 ,. . . Q_k. And we are given another query Q. Will we be able to compute Q only using the available views?

The above question, calling it “the question of determinacy”, sounds almost philosophical. One can easily imagine a bearded man in himation chained to the wall of a cave watching the views projected on the wall and pondering whether, from what he is able to see, the reality can be faithfully reconstructed.Read More