publications

2024

1. arXiv

   Learning in Hybrid Active Inference Models

   Poppy Collis, Ryan Singh, Paul F Kinghorn, and 1 more author

   2024
2. arXiv

   Only Strict Saddles in the Energy Landscape of Predictive Coding Networks?

   Francesco Innocenti, El Mehdi Achour, Ryan Singh, and 1 more author

   2024
3. arXiv

   Hybrid Recurrent Models Support Emergent Descriptions for Hierarchical Planning and Control

   Poppy Collis, Ryan Singh, Paul Kinghorn, and 1 more author

   In ICML 2024 Workshop: Foundations of Reinforcement Learning and Control – Connections and Perspectives, 2024
4. CCIS

   Relative Representations for Cognitive Graphs

   Alex B. Kiefer, and Christopher L. Buckley

   In Active Inference, 2024

   Abs

   Although the latent spaces learned by distinct neural networks are not generally directly comparable, even when model architecture and training data are held fixed, recent work in machine learning [13] has shown that it is possible to use the similarities and differences among latent space vectors to derive “relative representations” with comparable representational power to their “absolute” counterparts, and which are nearly identical across models trained on similar data distributions. Apart from their intrinsic interest in revealing the underlying structure of learned latent spaces, relative representations are useful to compare representations across networks as a generic proxy for convergence, and for zero-shot model stitching [13].

5. CCIS

   Exploring Action-Centric Representations Through the Lens of Rate-Distortion Theory

   Miguel De Llanza Varona, Christopher Buckley, and Beren Millidge

   In Active Inference, 2024

   Abs

   Organisms have to keep track of the information in the environment that is relevant for adaptive behaviour. Transmitting information in an economical and efficient way becomes crucial for limited-resourced agents living in high-dimensional environments. The efficient coding hypothesis claims that organisms seek to maximize the information about the sensory input in an efficient manner. Under Bayesian inference, this means that the role of the brain is to efficiently allocate resources in order to make predictions about the hidden states that cause sensory data. However, neither of those frameworks accounts for how that information is exploited downstream, leaving aside the action-oriented role of the perceptual system. Rate-distortion theory, which defines optimal lossy compression under constraints, has gained attention as a formal framework to explore goal-oriented efficient coding. In this work, we explore action-centric representations in the context of rate-distortion theory. We also provide a mathematical definition of abstractions and we argue that, as a summary of the relevant details, they can be used to fix the content of action-centric representations. We model action-centric representations using VAEs and we find that such representations i) are efficient lossy compressions of the data; ii) capture the task-dependent invariances necessary to achieve successful behaviour; and iii) are not in service of reconstructing the data. Thus, we conclude that full reconstruction of the data is rarely needed to achieve optimal behaviour, consistent with a teleological approach to perception.

6. CCIS

   Understanding Tool Discovery and Tool Innovation Using Active Inference

   Poppy Collis, Paul F. Kinghorn, and Christopher L. Buckley

   In Active Inference, 2024

   Abs

   The ability to invent new tools has been identified as an important facet of our ability as a species to problem solve in dynamic and novel environments [17]. While the use of tools by artificial agents presents a challenging task and has been widely identified as a key goal in the field of autonomous robotics, far less research has tackled the invention of new tools by agents. In this paper, (1) we articulate the distinction between tool discovery and tool innovation by providing a minimal description of the two concepts under the formalism of active inference. We then (2) apply this description to construct a toy model of tool innovation by introducing the notion of tool affordances into the hidden states of the agent’s probabilistic generative model. This particular state factorisation facilitates the ability to not just discover tools but invent them through the offline induction of an appropriate tool property. We discuss the implications of these preliminary results and outline future directions of research.

2023

1. arXiv

   Brain-Inspired Computational Intelligence via Predictive Coding

   Tommaso Salvatori, Ankur Mali, Christopher L. Buckley, and 4 more authors

   2023
2. ICML

   Attention as Implicit Structural Inference

   Ryan Singh, and Christopher Buckley

   In ICML 2023 Workshop on Structured Probabilistic Inference & Generative Modeling, 2023
3. ICML

   Understanding Predictive Coding as a Second-Order Trust-Region Method

   Francesco Innocenti, Ryan Singh, and Christopher Buckley

   In ICML Workshop on Localized Learning (LLW), 2023
4. arXiv

   Attention: Marginal Probability is All You Need?

   Ryan Singh, and Christopher L. Buckley

   2023
5. PoL

   From the free energy principle to a confederation of Bayesian mechanics: Reply to comments on “How particular is the physics of the free energy principle?”

   Miguel Aguilera, Beren Millidge, Alexander Tschantz, and 1 more author

   Physics of Life Reviews, 2023
6. PLoS Comp. Bio.

   A behavioral and modeling study of control algorithms underlying the translational optomotor response in larval zebrafish with implications for neural circuit function

   John G Holman, Winnie W K Lai, Paul Pichler, and 3 more authors

   PLoS Computational Biology, 2023
7. CCIS

   A Message Passing Perspective on Planning Under Active Inference

   Magnus Koudahl, Christopher L. Buckley, and Bert Vries

   In Active Inference, 2023
8. CCIS

   Active Inference Successor Representations

   Beren Millidge, and Christopher L. Buckley

   In Active Inference, 2023
9. CCIS

   Capsule Networks as Generative Models

   Alex B. Kiefer, Beren Millidge, Alexander Tschantz, and 1 more author

   In Active Inference, 2023
10. CCIS

    Knitting a Markov Blanket is Hard When You are Out-of-Equilibrium: Two Examples in Canonical Nonequilibrium Models

    Miguel Aguilera, Ángel Poc-López, Conor Heins, and 1 more author

    In Active Inference, 2023
11. CCIS

    Preventing Deterioration of Classification Accuracy in Predictive Coding Networks

    Paul F. Kinghorn, Beren Millidge, and Christopher L. Buckley

    In Active Inference, 2023
12. CCIS

    Spin Glass Systems as Collective Active Inference

    Conor Heins, Brennan Klein, Daphne Demekas, and 2 more authors

    In Active Inference, 2023

2022

1. Neural Comput.

   Predictive Coding Approximates Backprop Along Arbitrary Computation Graphs

   Beren Millidge, Alexander Tschantz, and Christopher L. Buckley

   Neural Computation, May 2022
2. arXiv

   Hybrid Predictive Coding: Inferring, Fast and Slow

   Alexander Tschantz, Beren Millidge, Anil K Seth, and 1 more author

   May 2022