We introduce Active Predictive Coding Networks (APCNs), a new class of neural networks that solve a major problem posed by Hinton and others in the fields of artificial intelligence and brain modeling: how can neural networks learn intrinsic reference frames for objects and parse visual scenes into part-whole hierarchies by dynamically allocating nodes in a parse tree? APCNs address this problem by using a novel combination of ideas: (1) hypernetworks are used for dynamically generating recurrent neural networks that predict parts and their locations within intrinsic reference frames conditioned on higher objectlevel embedding vectors, and (2) reinforcement learning is used in conjunction with backpropagation for end-to-end learning of model parameters. The APCN architecture lends itself naturally to multilevel hierarchical learning and is closely related to predictive coding models of cortical function. Using the MNIST, Fashion-MNIST and Omniglot datasets, we demonstrate that APCNs can (a) learn to parse images into part-whole hierarchies, (b) learn compositional representations, and (c) transfer their knowledge to unseen classes of objects. With their ability to dynamically generate parse trees with part locations for objects, APCNs offer a new framework for explainable AI that leverages advances in deep learning while retaining interpretability and compositionality.
We introduce Active Predictive Coding Networks (APCNs), a new class of neural networks that solve a major problem posed by Hinton and others in the fields of artificial intelligence and brain modeling: how can neural networks learn intrinsic reference frames for objects and parse visual scenes into part-whole hierarchies by dynamically allocating nodes in a parse tree? APCNs address this problem by using a novel combination of ideas: (1) hypernetworks are used for dynamically generating recurrent neural networks that predict parts and their locations within intrinsic reference frames conditioned on higher object-level embedding vectors, and (2) reinforcement learning is used in conjunction with backpropagation for end-to-end learning of model parameters. The APCN architecture lends itself naturally to multilevel hierarchical learning and is closely related to predictive coding models of cortical function. Using the MNIST, Fashion-MNIST and Omniglot datasets, we demonstrate that APCNs can (a) learn to parse images into part-whole hierarchies, (b) learn compositional representations, and (c) transfer their knowledge to unseen classes of objects. With their ability to dynamically generate parse trees with part locations for objects, APCNs offer a new framework for explainable AI that leverages advances in deep learning while retaining interpretability and compositionality.
The original predictive coding model of Rao & Ballard (1999) focused on spatial prediction to explain spatial receptive fields and contextual effects in the visual cortex. Here, we introduce a new dynamic predictive coding model that achieves spatiotemporal prediction of complex natural image sequences using time-varying transition matrices. We overcome the limitations of static linear transition models (as in, e.g., Kalman filters) using a hypernetwork to adjust the transition matrix dynamically for every time step, allowing the model to predict using a time-varying mixture of possible transition dynamics. We developed a single level model with recurrent modulation of transition weights by a hypernetwork and a two-level hierarchical model with top-down modulation based on a hypernetwork. At each time step, the model predicts the next input and estimates a sparse neural code by minimizing prediction error. When exposed to natural movies, the model learned localized, oriented spatial filters as well as both separable and inseparable (direction-selective) space-time receptive fields at the first level, similar to those found in the primary visual cortex (V1). Longer timescale responses and stability at the second level also emerged naturally from minimizing prediction errors for the first level dynamics. Our results suggest that the multiscale temporal response properties of cortical neurons could be the result of the cortex learning a hierarchical generative model of the visual world with higher order areas predicting the transition dynamics of lower order areas.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.