Predictive Coding Model Detects Novelty on Different Levels of Representation Hierarchy.

Li TE
Tang M

It probably happens to you that you walk on the street, you see someone, and you feel this person is familiar even though you cannot recall who this person is, or where and when you saw them. Our brain has remarkable capacity for discrimination between novel and familiar stimuli. This paper proposes a computational model describing how brain networks determine if a stimulus is novel or familiar.

Scientific Abstract

Novelty detection, also known as familiarity discrimination or recognition memory, refers to the ability to distinguish whether a stimulus has been seen before. It has been hypothesized that novelty detection can naturally arise within networks that store memory or learn efficient neural representation because these networks already store information on familiar stimuli. However, existing computational models supporting this idea have yet to reproduce the high capacity of human recognition memory, leaving the hypothesis in question. This article demonstrates that predictive coding, an established model previously shown to effectively support representation learning and memory, can also naturally discriminate novelty with high capacity. The predictive coding model includes neurons encoding prediction errors, and we show that these neurons produce higher activity for novel stimuli, so that the novelty can be decoded from their activity. Additionally, hierarchical predictive coding networks detect novelty at different levels of abstraction within the hierarchy, from low-level sensory features like arrangements of pixels to high-level semantic features like object identities. Overall, based on predictive coding, this article establishes a unified framework that brings together novelty detection, associative memory, and representation learning, demonstrating that a single model can capture these various cognitive functions.

Energy of novel and familiar stimuli
A class of computational models of memory in the brain suggests that neuronal networks minimize an abstract function called energy, and memorized items have lower energy. Hence by evaluating an energy of a network after presentation of a stimulus, it is possible to discriminate if the stimulus has been seen before.
Citation

2025. Neural Comput, 37(8):1373-1408.

DOI
10.1162/neco_a_01769
Related Content
Paper
Author
Salvatori T
Song Y
Xu Z
Lukasiewicz T
In Proceedings of the 36th AAAI Conference on Artificial Intelligence‚ AAAI 2022 ‚Vancouver, BC, Canada, February 22--March 1‚ 2022 (Vol. 10177, pp. 507-524). AAAI Press.
Paper
Author
Song Y
Lukasiewicz T
Xu Z
2020. Adv Neural Inf Process Syst, 33:22566-22579.
Paper
Author
Möller M
Manohar S

2022. PLoS Comput Biol, 18(5)e1009816.

Paper
Author
Koolschijn RS
Shpektor A
Clarke WT
Ip IB
Emir UE

2021. eLife, 10:e70071

Paper
Author
Salvatori T
Song Y
Yordanov Y
Millidge B
Sha L
Emde C
Xu Z
Lukasiewicz T
2024. The Twelveth International Conference on Learning Representations