Abstract
The theoretical properties of active inference agents are impressive, but how do we realize effective agents in working hardware and software on edge devices? This is an interesting problem because the computational load for policy exploration explodes exponentially, while the computational resources are very limited for edge devices. In this paper, we discuss the necessary features for a software toolbox that supports a competent non-expert engineer to develop working active inference agents. We introduce a toolbox-in-progress that aims to accelerate the democratization of active inference agents in a similar way as TensorFlow propelled applications of deep learning technology.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Similar content being viewed by others
Notes
- 1.
For reference, we use the following abbreviations in this paper: Active Inference (AIF), Constrained Bethe Free Energy (CBFE), Expected Free Energy (EFE), (variational) Free Energy (FE), Free Energy Principle (FEP), Free Energy Minimization (FEM), Message Passing (MP), Reactive Message Passing (RMP).
- 2.
The computational load and complexity can only be equated in the absence of a Von Neumann bottleneck (i.e., with mortal computation or in-memory processing). This is because energy and time are ‘wasted’ by reading and writing to memory.
References
Akbayrak, S., Bocharov, I., de Vries, B.: Extended variational message passing for automated approximate Bayesian inference. Entropy 23(7), 815 (2021). ISSN: 1099-4300. https://doi.org/10.3390/e23070815. https://www.mdpi.com/1099-4300/23/7/815. Accessed 26 May 2023
Bagaev, D., de Vries, B.: Reactive message passing for scalable bayesian inference. Sci. Program. 2023 (2023). ISSN: 1058-9244. https://doi.org/10.1155/2023/6601690. https://www.hindawi.com/journals/sp/2023/6601690/. Accessed 28 May 2023
Beckers, J., et al.: Principled Pruning of Bayesian Neural Networks through Variational Free Energy Minimization (2022). https://doi.org/10.48550/ar**v.2210.09134. ar**v: 2210.09134. http://arxiv.org/abs/2210.09134. Accessed 26 May 2023
Bezanson, J., et al.: Julia: a fresh approach to numerical computing. SIAM Rev. 59(1), 65–98 (2017). ISSN: 0036-1445. https://doi.org/10.1137/141000671. https://epubs.siam.org/doi/10.1137/141000671. Accessed 03 Feb 2022
Champion, T., Grze, M., Bowman, H.: Realizing active inference in variational message passing: the outcome-blind certainty seeker. Neural Comput. 33(10), 2762–2826 (2021). ISSN: 0899-7667. https://doi.org/10.1162/neco_a_01422. Accessed 26 May 2023
Cox, M., van de Laar, T., de Vries, B.: A factor graph approach to automated design of Bayesian signal processing algorithms. Int. J. Approx. Reason. 104, 185–204 (2019). ISSN: 0888-613X. https://doi.org/10.1016/j.ijar.2018.11.002. http://www.sciencedirect.com/science/article/pii/S0888613X18304298. Accessed 16 Nov 2018
Distributive property. Wikipedia. Page Version ID: 1124679546 (2022). https://en.wikipedia.org/w/index.php?title=Distributive_property&oldid=1124679546. Accessed 26 May 2023
Friston, K., Parr, T., Zeidman, P.: Bayesian model reduction. ar**v:1805.07092 (2018). http://arxiv.org/abs/1805.07092. Accessed 28 May 2018
Friston, K., et al.: Active inference and epistemic value. Cogn. Neurosci. (2015). ISSN: 1758-8928. https://doi.org/10.1080/17588928.2015.1020053. Accessed 22 Feb 2015
Friston, K., et al.: Sophisticated inference. Neural Comput. 33(3), 713–763 (2021). ISSN 0899-7667. https://doi.org/10.1162/neco_a_01351. Accessed 14 Feb 2022
Friston, K.J., et al.: Designing Ecosystems of Intelligence from First Principles (2022). https://doi.org/10.48550/ar**v.2212.01354. ar**v:2212.01354. http://arxiv.org/abs/2212.01354. Accessed 08 Dec 2022
Friston, K.J., et al.: SPM12 toolbox (2014). http://www.fil.ion.ucl.ac.uk/spm/software/
Heins, C., et al.: pymdp: a Python library for active inference in discrete state spaces. ar**v:2201.03904 (2022). http://arxiv.org/abs/2201.03904. Accessed 03 Feb 2022
Lanczos, C.: The Variational Principles of Mechanics, 4th Revised edition, 464 p. Dover Publications, New York (1986). ISBN 978-0-486-65067-8
Loeliger, H.-A., et al.: On sparsity by NUV-EM, Gaussian message passing, and Kalman smoothing. In: 2016 Information Theory and Applications Workshop (ITA). 2016 Information Theory and Applications (ITA), La Jolla, CA, USA, pp. 1–10. IEEE (2016). ISBN: 978-1-5090-2529-9. https://doi.org/10.1109/ITA.2016.7888168. http://ieeexplore.ieee.org/document/7888168/. Accessed 21 July 2021
Senöz, I., et al.: Variational message passing and local constraint manipulation in factor graphs. Entropy 23(7), 807 (2021). ISSN: 1099-4300. https://doi.org/10.3390/e23070807
Smirnova, L., et al.: Organoid intelligence (OI): the new frontier in biocomputing and intelligence-in-a-dish. Front. Sci. (2023)
Acknowledgments
I would like to acknowledge my colleagues at BIASlab (http://biaslab.org) for the stimulating work environment and the anonymous reviewers for excellent feedback on the draft version. Some wording in this document, such as footnote (see footnote 2), comes straight from a reviewer.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2024 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
de Vries, B. (2024). Toward Design of Synthetic Active Inference Agents by Mere Mortals. In: Buckley, C.L., et al. Active Inference. IWAI 2023. Communications in Computer and Information Science, vol 1915. Springer, Cham. https://doi.org/10.1007/978-3-031-47958-8_11
Download citation
DOI: https://doi.org/10.1007/978-3-031-47958-8_11
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-47957-1
Online ISBN: 978-3-031-47958-8
eBook Packages: Computer ScienceComputer Science (R0)