# ISAAC-IAS Research reading group on “Statistical theory and methods for embodied, enactive and ecological cognitive science”

**1. Motivation**

For the last three decades, embodied approaches to life and cognition have relied heavily in concepts from nonlinear dynamical systems theory for describing concepts about adaptation, emergence and self-organization. Some examples are modelling approaches in evolutionary robotics or dynamical or dynamical accounts of perception and learning in ecological psychology. During this time, dynamical systems approaches have been successful in providing the mathematical scaffolding to conceptualize the complexity of biological systems without relying on notions of representation or information processing in a computational sense. Nevertheless, dynamical systems approaches have also reached some important limits regarding the scalability of models, or the quantification of important emergent properties a system.

In contrast, some of this shortcomings are being faced in different fields of complex systems research by complementing dynamical systems approaches with statistical tools borrowed from information theory, bayesian inference, or statistical mechanics. To name a few, combination of techniques from these fields allows to infer better dynamical models from experimental data (e.g. using machine learning and inference techniques to model neural networks that map the activity of observed neurons), quantify important aspects of cognitive processes (e.g. information theoretical tools to measure causal relations, closure or autonomy in complex systems), or develop formal principles connecting different phenomena under a common framework (e.g. information maximization or bayesian inference principles, etc.).

Following the tradition of 1960 cyberneticians, this reading seminar series aims to join points of view from information theory, computer science and neuroscience into current problems in embodied cognitive science. In this line, the group sessions will focus in the study of theory and techniques from different fields: information theory, machine learning, variational bayesian methods, statistical mechanics, etc. with the goal to discuss their applicability in the context of embodied, enactive and ecological cognition.

**2. Objectives**

- Introduce theoretical and technical tools for the study the statistical behaviour of complex dynamical systems
- Facilitate for participants an awareness of basic items of literature and shared background of knowledge.
- Discuss the application of these methods to current problems in embodied, enactive and ecological cognition.
- Debate the underlying assumptions of different theoretical frameworks using statistical tools, in order to delineate how a statistical approach to embodied and ecological problems might look like.

**3. Orientation**

The reading sessions are oriented as working sessions in collaboration between members of the ISAAC lab a the University of Zaragoza and the IAS Research at the University of the Basque Country. The sessions are open to other participants, although the readings require basic familiarity with calculus, probability theory, and linear algebra as taught in a first-year undergraduate course on mathematics for scientists and engineers. Sessions marked with an asterisk (∗) will be designed to be open to different profiles and mathematical knowledge will not be a limitation.

**4. Format**

Twelve fortnightly reading seminars lasting 2 hours around the reading of chapters of the following introductory texts to different topics about information theory, probabilistic methods and inference:

- MacKay, D. J. (2003). Information theory, inference and learning algorithms. Cambridge university press.
- Bishop, C. M. (2012). Pattern recognition and machine learning. Springer.

Seminars will be run by the reading group coordinator and a rotating member of the group which will present a short summary of the session’s readings and a set of questions for the following debate. Participants will be given pieces of reading prior to the seminar, which will serve to support the discussion of the theoretical and technical tools covered in the session.

Special sessions will be organized around philosophical debates related with information theory, ecological information, autonomy and active inference. These sessions will be focused on the philosophical debates rather than in the mathematical notions. In these special sessions the following texts will be discussed:

- Krippendorff, K. (2009). Ross Ashby’s information theory: a bit of history, some solutions to problems, and what we face today.
*International journal of general systems*,*38*(2), 189-212. - Gibson, J. J. (2014).
*The ecological approach to visual perception: classic edition*. Psychology Press (fragment) - Bertschinger, N., Olbrich, E., Ay, N., & Jost, J. (2008). Autonomy: An information theoretic perspective.
*Biosystems*,*91*(2), 331-345. - Friston, K., Mattout, J., & Kilner, J. (2011). Action understanding and active inference.
*Biological cybernetics*,*104*(1-2), 137-160.

**5. Schedule, topics, and readings**

Seminars will take place on alternate Wednesdays from 15:00 to 17:00h, open to online and in-person participation at the ISAAC and IAS Research labs.

Week 1: Introduction (MacKay, Chapters 1-3) – 03/10

Week 2: Data Compression (MacKay, Chapters 4-7) – 17/10

Week 3 (∗) : Ashby’s Information Theory (Krippendorff, 2009) – 31/10

Week 4: Noisy Channel-Coding (MacKay, Chapters 8-11) – 14/11

Week 5: Information Theory I (MacKay, Chapters 12-14) – 28/11

Week 6: Information Theory II (MacKay, Chapters 15-19) – 12/12

Week 7 (∗) : Ecological information and information theory (Gibson, 2014) – 09/01

Week 8: Inference (MacKay, Chapters 20-27) – 23/01

Week 9: Monte Carlo Methods and Ising models (MacKay, Chapters 28-32) – 06/02

Week 10 (∗) : Autonomy and Information Theory (Bertschinger et al, 2008) – 20/02

Week 11: Variational and Bayesian methods I (MacKay, Chapters 33-37) – 06/03

Week 12: Variational and Bayesian methods II (Bishop, Chapters 9-10) – 20/03

Week 13 (∗) : Active Inference (Friston et al, 2011) – 03/04

Week 14: Neural Networks I (MacKay, Chapters 38-46) – 17/04

Week 15: Neural Networks II (Bishop, Chapters 5 and 8) – 30/04

Week 16: Sparse Graph Codes (MacKay, Chapters 47-50) -15/05

Sessions marked with an asterisk (∗) will focus more on the philosophical discussion rather than in mathematical concepts

**6. Coordination and more information**

In order to join the reading group or request further information, please contact the coordinator:

Miguel Aguilera