Pattern Mining with Paspailleur: Intro to a new Python package

Room TBD

Pattern Mining (also known as Data Mining) is a field of science designed for extracting symbolic patterns in datasets. The mined patterns can then be used for constructing implication systems, for mining association rules, and for conceptual clustering and classification on the data. Pattern mining research is often focused on a specific type of complex data, such as itemsets (i.e. binary descriptions), numbers, sequences, and graphs. This talk will present the Pattern Structures framework that encompasses all listed data types with a single formalism.

For a long time, the research on generic Pattern Structures was mostly theoretical. However, this January, we have found a way to treat various types of data in a unified way and to make the computations work in a matter of seconds and minutes. The new approach – called Atomic Patterns – is implemented in Paspailleur Python package scheduled for release in May 2025.

The first part of the talk will present the various tasks of Pattern Mining and how they can be approached using Paspailleur package. Then, during the Q&A section, we will try to find an answer to the biggest question of my thesis, which is “How can Pattern Structures enhance the work of other researchers?” In particular, how can Pattern Mining and Paspailleur be used within the working groups of Knowledge Representation and Reasoning, Hybrid AI, Frugal AI, XAI, and LLMs and NLP.