Programme
Invited talks
Active Learning and Active Inference in Exploration Philipp Schwartenbeck
Successful behaviour depends on the right balance between maximising reward and soliciting information about the world. I will discuss how different types of information-gain emerge when casting behaviour as surprise minimisation and planning as an inferential process. This formulation provides two distinct mechanisms for goal-directed exploration that express separable profiles of active sampling to reduce uncertainty. 'Hidden state' exploration motivates agents to sample unambiguous observations to accurately infer the (hidden) state of the world. Conversely, 'model parameter' exploration, compels agents to sample outcomes associated with high uncertainty, if they are informative for their representation of the task structure. I will try to provide an introductory illustration of the emergence of these types of 'Bayes-optimal' exploratory behaviour, termed active inference and active learning, and discuss possible future developments and experimental investigations of such implementations in artificial and biological agents.
[Presentation]
Putting An End to End-to-End: Gradient-Isolated Learning of Representations Sindy Löwe
We propose a novel deep learning method for local self-supervised representation learning that does not require labels nor end-to-end backpropagation but exploits the natural order in data instead. Inspired by the observation that biological neural networks appear to learn without backpropagating a global error signal, we split a deep neural network into a stack of gradient-isolated modules. Each module is trained to maximally preserve the information of its inputs using the InfoNCE bound from Oord et al. [2018]. Despite this greedy training, we demonstrate that each module improves upon the output of its predecessor, and that the representations created by the top module yield highly competitive results on downstream classification tasks in the audio and visual domain. The proposal enables optimizing modules asynchronously, allowing large-scale distributed training of very deep neural networks on unlabelled datasets.
[Presentation]
The Free Energy Principle and Active Inference in silico and in vivo, visual sampling and 'world model' building Rosalyn Moran
The theory of Active Inference proposes that all biological agents retain self-ness by minimizing their long-term average surprisal. In information theoretic terms, Free Energy provides a soluble approximation to this long-term surprise 'now' and necessitates the development of a generative model of the environment within the agent itself. The minimization of this quantity via a gradient flow is purported to be the purpose of neuronal activity in the brain and thus provides a mapping from brain activity to their first-principle computations. In this talk I will outline the theory of Active Inference and describe how discrete and continuous-time systems that perceive and act can be built in silico, while providing evidence for these implementations in neurobiological and behavioral recordings. Using two experiments in human participants, I aim to demonstrate that human visual search and classification of the MNIST dataset (experiment 1) and world model building and adjustment in a maze task (experiment 2) can be cast as Active Inference processes that utilize neurobiologically plausible architectures comprising prediction in visual hierarchies and alterations in precision via neuromodulation.
[Presentation]
Accepted presentations
Confirmatory evidence that healthy individuals can adaptively adjust prior expectations and interoceptive precision estimates
Ryan Smith, Rayus Kuplicki, Adam Teed, Valerie Upshaw and Sahib S. Khalsa
[Presentation]
On the relationship between active inference and control as inference
Beren Millidge, Alexander Tschantz, Anil Seth and Christopher L. Buckley
[Presentation]
Visual search as active inference
Emmanuel Daucé and Laurent Perrinet.
[Presentation]
Integrated World Modeling Theory (IWMT) Implemented: Towards Reverse Engineering Consciousness with the Free Energy Principle and Active Inference
Adam Safron
[Presentation]
A deep active inference model of the rubber-hand illusion
Thomas Rood, Marcel van Gerven and Pablo Lanillos
[Presentation]
Sleep: Model Reduction in Deep Active Inference
Samuel Wauthier, Ozan Catal, Cedric De Boom, Tim Verbelen and Bart Dhoedt
[Presentation]
Active Inference for Fault Tolerant Control of Robot Manipulators with Sensory Faults
Corrado Pezzato, Mohamed Baioumy, Carlos Hernandez Corbato, Nick Hawes, Martijn Wisse and Riccardo Ferrari
[Presentation]
Modulation of viability signals for self-regulatory control
Alvaro Ovalle and Simon Lucas
[Presentation]
Active Inference or Control as Inference? A Unifying View
Abraham Imohiosen, Joe Watson and Jan Peters
[Presentation]
A Worked Example of Fokker-Planck based Active Inference
Magnus T. Koudahl and Bert de Vries
[Presentation]
You Only Look... as much as you have to: Using the Free Energy Principle for Active Vision
Toon Van de Maele, Tim Verbelen, Ozan Catal, Cedric De Boom and Bart Dhoedt
[Presentation]
Bayesian hyperparameter dynamics in a Markov chain
Martin Biehl and Ryota Kanai
[Presentation]
Deep active inference for Partially Observable MDPs
Otto van der Himst and Pablo Lanillos
[Presentation]
Accepted posters
Online system identification in a Duffing oscillator by free energy minimisation
Wouter Kouw
[Presentation] [Poster]
Causal blankets: Theory and algorithmic framework
Fernando E. Rosas, Pedro A.M. Mediano, Martin Biehl, Shamil Chandaria and Daniel Polani
[Presentation]
Sophisticated Affective Inference: Simulating Affective Dynamics Induced by Imagined Future Events
Casper Hesp, Alexander Tschantz, Beren Millidge, Maxwell Ramstead, Karl Friston and Ryan Smith
[Presentation]
Learning Where to Park
Burak Ergul, Thijs van de Laar, Magnus Koudahl, Martin Roa Villescas and Bert de Vries
[Presentation]
End-Effect Exploration Drive for Effective Motor Learning
Emmanuel Daucé.
[Presentation]
Hierarchical Gaussian filtering of sufficient statistic time series for active inference
Christoph Mathys and Lilian A.E. Weber
[Presentation]