Bidirectional predictive coding

2025-05-29

AI Summary

Traditional predictive coding (PC) models support only unidirectional (generative or discriminative) inference, failing to emulate the parallel bidirectional information flow observed in the visual cortex and thus exhibiting limited performance in multimodal learning and missing-data inference. To address this, we propose bidirectional predictive coding (bPC), the first biologically plausible hierarchical recurrent neural circuit that jointly implements top-down (generative) and bottom-up (discriminative) inference via a unified energy function optimized end-to-end. Grounded in predictive coding theory, variational inference, and neural dynamics modeling, bPC enables bidirectional error propagation and state updates. Experiments demonstrate that bPC matches or surpasses unidirectional PC models on standalone generative and discriminative tasks, while achieving substantial gains in multimodal fusion and missing-modality imputation—aligning more closely with neurobiological principles of visual inference.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Multi-modal Vision
• Machine Learning: Probabilistic Circuits and Graphical Models

Application Category

• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

Predictive coding (PC) is an influential computational model of visual learning and inference in the brain. Classical PC was proposed as a top-down generative model, where the brain actively predicts upcoming visual inputs, and inference minimises the prediction errors. Recent studies have also shown that PC can be formulated as a discriminative model, where sensory inputs predict neural activities in a feedforward manner. However, experimental evidence suggests that the brain employs both generative and discriminative inference, while unidirectional PC models show degraded performance in tasks requiring bidirectional processing. In this work, we propose bidirectional PC (bPC), a PC model that incorporates both generative and discriminative inference while maintaining a biologically plausible circuit implementation. We show that bPC matches or outperforms unidirectional models in their specialised generative or discriminative tasks, by developing an energy landscape that simultaneously suits both tasks. We also demonstrate bPC's superior performance in two biologically relevant tasks including multimodal learning and inference with missing information, suggesting that bPC resembles biological visual inference more closely.

Problem

• Bidirectional predictive coding combines generative and discriminative inference
• Unidirectional models underperform in bidirectional processing tasks
• bPC improves multimodal learning and inference with missing data

Innovation

• Bidirectional predictive coding combines generative and discriminative models
• Energy landscape optimised for dual-task performance
• Biologically plausible circuit enhances multimodal learning