Neuromorphic Computing with Multi-Frequency Oscillations: A Bio-Inspired Approach to Artificial Intelligence

TL;DR

This work addresses the gap between artificial and biological intelligence by introducing a tripartite brain-inspired architecture with functionally specialized perceptual, auxiliary, and executive systems. Temporal coordination is achieved through multi-frequency neural oscillations, neuromodulation, and a certainty measure, enabling dynamic, context-sensitive processing across modules. Empirical results on visual benchmarks show up to ~2.18% accuracy gains and substantial reductions in computation iterations, along with improved alignment to human categorization patterns and robustness to noise. The framework provides a theoretical and practical foundation for brain-like AI and offers a pathway to extend these principles to other cognitive domains such as language and reasoning.

Abstract

Despite remarkable capabilities, artificial neural networks exhibit limited flexible, generalizable intelligence. This limitation stems from their fundamental divergence from biological cognition that overlooks both neural regions' functional specialization and the temporal dynamics critical for coordinating these specialized systems. We propose a tripartite brain-inspired architecture comprising functionally specialized perceptual, auxiliary, and executive systems. Moreover, the integration of temporal dynamics through the simulation of multi-frequency neural oscillation and synaptic dynamic adaptation mechanisms enhances the architecture, thereby enabling more flexible and efficient artificial cognition. Initial evaluations demonstrate superior performance compared to state-of-the-art temporal processing approaches, with 2.18\% accuracy improvements while reducing required computation iterations by 48.44\%, and achieving higher correlation with human confidence patterns. Though currently demonstrated on visual processing tasks, this architecture establishes a theoretical foundation for brain-like intelligence across cognitive domains, potentially bridging the gap between artificial and biological intelligence.

Figures (4)

• Figure 1: Tripartite Brain Cognitive Architecture in the Human Brain. In visual tasks, three functionally specialized systems Perceptual (visual cortex and temporal lobe), Auxiliary (ventral tegmental area), and Executive (frontal lobe) reflect the regional collaboration of biological neural organization. Visual semantics from perceptual feature processing, while synergistic interactions between frontal lobe synaptic adaptation and ventral tegmental neuromodulation collectively support categorical decision-making.
• Figure 2: Tripartite Brain-Inspired Architecture with neural oscillation and adaptive processing. a, Information flow from visual input through Perceptual (visual cortex and temporal lobe), Auxiliary (ventral tegmental area), and Executive (frontal lobe) systems, integrating hierarchical processing with neural dynamics for categorical output. b, Neural oscillation mechanisms with frequency band assignment ($\gamma$, $\beta$, $\alpha$, $\theta$). c, Neuromodulation network that adjusts oscillatory parameters based on context. d, Adaptive processing showing complexity-based synaptic pathway selection. e, Iterative control mechanism that dynamically determines computation termination based on certainty metrics.
• Figure 3: Robustness to input noise and human-alignment analysis. a-b, Accuracy and average iteration steps under different fixed gaussian noise levels ($\sigma$) for CIFAR10. c, Comparison between CTM and human categorization on CIFAR-10H. d, Comparison between our model and human categorization on CIFAR-10H.
• Figure 4: Ablation studies and visualization of Tripartite Brain-Inspired Architecture. a, Impact of neural oscillation and modulation on performance and efficiency. b, Effect of synaptic complexity on accuracy and computational iterations. c, Accuracy improvement with increasing neuron density. d, Computational efficiency gains with increasing neuron count. e, Attention map visualization across early, middle, and final processing stages. f, Phase coherence patterns showing characteristic changes throughout computation. g, Evolution of neural activation patterns during iterative processing.