Enhancing deep neural networks through complex-valued representations and Kuramoto synchronization dynamics

2025-02-28

AI Summary

Deep learning models struggle with object binding in multi-object visual scenes, failing to accurately distinguish and integrate features belonging to the same object. To address this, we propose a neuroscience-inspired, synchrony-driven object binding mechanism: for the first time, we embed Kuramoto phase synchronization dynamics into deep neural networks, leveraging complex-valued weights and learnable phase encodings to achieve differentiable, dynamic, object-level feature clustering via feature-phase alignment. Our method supports both feedforward and feedback loop architectures and introduces a phase-driven feature grouping module. Evaluated on multi-object classification tasks—including overlapping handwritten digits, noisy images, and out-of-distribution geometric transformations—our model significantly outperforms real-valued baselines and ablated complex-valued models without synchronization. Results demonstrate that phase synchronization substantially enhances robustness and generalization, validating its functional role in perceptual grouping.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Neural synchrony is hypothesized to play a crucial role in how the brain organizes visual scenes into structured representations, enabling the robust encoding of multiple objects within a scene. However, current deep learning models often struggle with object binding, limiting their ability to represent multiple objects effectively. Inspired by neuroscience, we investigate whether synchrony-based mechanisms can enhance object encoding in artificial models trained for visual categorization. Specifically, we combine complex-valued representations with Kuramoto dynamics to promote phase alignment, facilitating the grouping of features belonging to the same object. We evaluate two architectures employing synchrony: a feedforward model and a recurrent model with feedback connections to refine phase synchronization using top-down information. Both models outperform their real-valued counterparts and complex-valued models without Kuramoto synchronization on tasks involving multi-object images, such as overlapping handwritten digits, noisy inputs, and out-of-distribution transformations. Our findings highlight the potential of synchrony-driven mechanisms to enhance deep learning models, improving their performance, robustness, and generalization in complex visual categorization tasks.

Problem

• Enhancing object encoding in deep learning models using synchrony-based mechanisms.
• Improving visual categorization through complex-valued representations and Kuramoto dynamics.
• Addressing object binding challenges in multi-object image tasks.

Innovation

• Complex-valued representations enhance neural networks.
• Kuramoto dynamics promote phase alignment in models.
• Synchrony-driven mechanisms improve visual categorization tasks.