Event details from: Pointwise information decomposition for complex systems (hybrid)

Date
28 September 2023
Time
11:00 – 12:00
Room
Library

Title

Pointwise information decomposition for complex systems

Abstract

The aim of information decomposition is to provide a mathematical framework that partitions the total information provided by a set of source variables about a target variable into i) the unique information that is provided by each individual source variable, ii) the shared information that is redundantly provided by two or more source variables, and iii) the synergistic information that is only attainable from simultaneous knowledge of two or more source variables. Such a decomposition has many potential applications in the sciences: for instance, quantifying the synergistic interaction between multiple incoming neural stimuli that are fused to create some output signal, or for determining the extent to which a particular phenotypic trait depends uniquely on each individual source gene, redundantly on two or more genes or is determined by some synergistic combination of several genes. Most approaches to information decomposition focus on decomposition the average information provided the variables involved.

In this talk, I will discuss why this is unsatisfactory, especially when in comes to analysing complex systems. I will provide a brief overview of the pointwise perspective on information theory and then discuss the challenges associated with determining a pointwise information decomposition. I will then discuss the pointwise partial information decomposition that we introduced before closing by demonstrating how this theory can be used to quantify emergent intrinsic computation in canonical complex systems, namely elementary cellular automata.

If you wish to to attend this seminar online, please send an email to w.merbis@uva.nl to receive the zoom-link.