An active inference implementation of phototaxis

Baltieri, Manuel; Buckley, Christopher L.

doi:10.7551/ecal_a_011

Cited by 46 publications

(85 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In summary, active inference proposes that agents learn and update a probabilistic model of their world, and act to maximize the evidence for this model. However, in contrast to previous 'perception-oriented' approaches to constructing probabilistic models (Baltieri and Buckley, 2017), active inference requires an agent's model to be intrinsically biased towards certain (favourable) observations. The goal is not, therefore, to construct a model that accurately captures the true causal structure underlying observations, but is instead to learn a model that is tailored to a specific set of prior preferences, and thus tailored to a specific set of agent-environment interactions.…”

Section: Resultsmentioning

confidence: 99%

“…One approach to this problem is for organisms to selectively model their world in a way that supports action (Seth, 2015;Seth and Tsakiris, 2018;Baltieri and Buckley, 2017;Clark, 2015;Gibson, 2014). We refer to such models as action-oriented, as their functional purpose is to enable adaptive behaviour, rather than to represent the world in a complete or accurate manner.…”

Section: Introductionmentioning

confidence: 99%

“…An action-oriented representation of the world can depart from a veridical representation in a number of ways. First, because only a subset of the states and contingencies in an environment will be relevant for behaviour, action-oriented models need not exhaustively model their environment (Baltieri and Buckley, 2017). Moreover, specific misrepresentations may prove to be useful for action (Wiese, 2017;McKay and Dennett, 2009;Mendelovici, 2013;M.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Learning action-oriented models through active inference

Tschantz

Seth

Buckley

2019

Preprint

Self Cite

View full text Add to dashboard Cite

Converging theories suggest that organisms learn and exploit probabilistic models of their environment. However, it remains unclear how such models can be learned in practice. The open-ended complexity of natural environments means that it is generally infeasible for organisms to model their environment comprehensively. Alternatively, action-oriented models attempt to encode a parsimonious representation of adaptive agent-environment interactions. One approach to learning action-oriented models is to learn online in the presence of goal-directed behaviours. This constrains an agent to behaviourally relevant trajectories, reducing the diversity of the data a model need account for. Unfortunately, this approach can cause models to prematurely converge to sub-optimal solutions, through a process we refer to as a bad-bootstrap. Here, we exploit the normative framework of active inference to show that efficient action-oriented models can be learned by balancing goal-oriented and epistemic (information-seeking) behaviours in a principled manner. We illustrate our approach using a simple agent-based model of bacterial chemotaxis. We first demonstrate that learning via goal-directed behaviour indeed constrains models to behaviorally relevant aspects of the environment, but that this approach is prone to sub-optimal convergence. We then demonstrate that epistemic behaviours facilitate the construction of accurate and comprehensive models, but that these models are not tailored to any specific behavioural niche and are therefore less efficient in their use of data. Finally, we show that active inference agents learn models that are parsimonious, tailored to action, and which avoid bad bootstraps and sub-optimal convergence. Critically, our results indicate that models learned through active inference can support adaptive behaviour in spite of, and indeed because of, their departure from veridical representations of the environment. Our approach provides a principled method for learning adaptive models from limited interactions with an environment, highlighting a route to sample efficient learning algorithms.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Learning action-oriented models through active inference

Tschantz

Seth

Buckley

2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…step disturbances and more in general noise, given by their filtering properties (I term), and 2) an only approximate model of the dynamics of the process to regulate, based on a linearisation around the target state. While this might look incompatible with work in active inference formulations suggesting a link to optimal control strategies with perfect models of process/environment, we argued previously that this needs not be the case [6]. One of the main strengths of active inference lies, according to us, in its general formulation and in generative models that do not have to mirror the dynamics of the environment, perhaps due to limitations or constraints of a system (e.g.…”

Section: Pid Controlmentioning

confidence: 86%

“…the dynamics of the world the agent interacts with. In recent work we have argued that this needs not be the case [6], especially if we consider simple living systems with limited resources. We intuitively don't expect an ant to explicitly model the environment where it forages, performing complex simulations of the environment in its brain (cf.…”

Section: Active Inference and Controlmentioning

confidence: 99%

A probabilistic interpretation of PID controllers using active inference

Baltieri

Buckley

2018

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. In the past few decades, probabilistic interpretations of brain functions have become widespread in cognitive science and neuroscience. The Bayesian brain hypothesis, predictive coding, the free energy principle and active inference are increasingly popular theories of cognitive functions that claim to unify understandings of life and cognition within general mathematical frameworks derived from information theory, statistical physics and machine learning. Furthermore, it has been argued that one such proposal, active inference, combines both information and control theory and has its roots in cybernetics studies of the brain. The connections between information and control theory have been discussed since the 1950's by scientists like Shannon and Kalman and have recently risen to prominence in modern stochastic optimal control theory. However, the implications of the confluence of these two theoretical frameworks for the biological sciences have been slow to emerge. Here we argue that if the active inference proposal is to be taken as a general process theory for biological systems, we need to consider how existing control theoretical approaches to biological systems relate to it. In this work we will focus on PID (Proportional-Integral-Derivative) controllers, one of the most common types of regulators employed in engineering and more recently used to explain behaviour in biological systems, e.g. chemotaxis in bacteria and amoebae or robust adaptation in biochemical networks. Using active inference, we derive a probabilistic interpretation of PID controllers, showing how they can fit a more general theory of life and cognition under the principle of (variational) free energy minimisation once we use only simple linear generative models.

show abstract

Hybrid Life: Integrating biological, artificial, and cognitive systems

Baltieri

Iizuka

Witkowski

et al. 2023

WIRES Cognitive Science

Self Cite

View full text Add to dashboard Cite

Artificial life is a research field studying what processes and properties define life, based on a multidisciplinary approach spanning the physical, natural, and computational sciences. Artificial life aims to foster a comprehensive study of life beyond “life as we know it” and toward “life as it could be,” with theoretical, synthetic, and empirical models of the fundamental properties of living systems. While still a relatively young field, artificial life has flourished as an environment for researchers with different backgrounds, welcoming ideas, and contributions from a wide range of subjects. Hybrid Life brings our attention to some of the most recent developments within the artificial life community, rooted in more traditional artificial life studies but looking at new challenges emerging from interactions with other fields. Hybrid Life aims to cover studies that can lead to an understanding, from first principles, of what systems are and how biological and artificial systems can interact and integrate to form new kinds of hybrid (living) systems, individuals, and societies. To do so, it focuses on three complementary perspectives: theories of systems and agents, hybrid augmentation, and hybrid interaction. Theories of systems and agents are used to define systems, how they differ (e.g., biological or artificial, autonomous, or nonautonomous), and how multiple systems relate in order to form new hybrid systems. Hybrid augmentation focuses on implementations of systems so tightly connected that they act as a single, integrated one. Hybrid interaction is centered around interactions within a heterogeneous group of distinct living and nonliving systems. After discussing some of the major sources of inspiration for these themes, we will focus on an overview of the works that appeared in Hybrid Life special sessions, hosted by the annual Artificial Life Conference between 2018 and 2022.This article is categorized under: Neuroscience > Cognition Philosophy > Artificial Intelligence Computer Science and Robotics > Robotics

show abstract

An active inference implementation of phototaxis

Cited by 46 publications

References 22 publications

Learning action-oriented models through active inference

Learning action-oriented models through active inference

A probabilistic interpretation of PID controllers using active inference

Hybrid Life: Integrating biological, artificial, and cognitive systems

Contact Info

Product

Resources

About