Workshop on “Integrative Machine learning”

Satellite Workshop at the

6th International Conference on Machine Learning, Optimization & Data Science

July 18, 2020 – Certosa di Pontignano, Siena, Italy

Conference Room: Bracci

Full-Day Workshop @ LOD 2020

Integrative Machine Learning

Machine learning is very effective at jointly learning feature representations and classification models, especially when dealing with high dimensional input patterns. Machine Learning solutions tend to be vertical, in the sense that they can well solve the specific task they are trained for, but the lack of emergence of a general and portable intelligence is nowadays seen as major limitation. Furthermore, machine learning is still limited when there is the need for consistent and robust decisions in complex environments, where the learning data can inevitably cover a small portion of all possible variants. The integration of machine learning and logic reasoning is an open-research problem, which could overtake these fundamental limitations, leading to the development of real intelligent agents.

This workshop covers topics in “Integrative Machine Learning”, where Machine Learning is augmented with knowledge representation and logic reasoning. In particular, we will discuss the different trade-offs on the application of constraint-based vs probabilistic solutions to represent the knowledge. Other open research problems will be discussed like: are directed or undirected models more suitable for Integrative ML? How to allow a flexible reasoning process without limiting the scalability of machine learning solutions? How to get advantage of modern tensor-based computational frameworks within integrative ML? Is an induction/deduction loop possible in machine learning like it happens in the development of human cognition?

Main topics

Integration of Machine Learning with

– Probabilistic Logic Programming

– Statistical Relational Learning

– Integer Programming and Optimization

– Information-based Principles

– Fuzzy Logic and Reasoning

– Constraint-based learning

– Graphical Models

– Knowledge Extraction and Representation