Understanding Sleep Mechanisms in Brain Through Quantum-inspired Spiking Neural Networks

The human brain achieves something computationally remarkable: it continuously transforms overwhelming sensory complexity into coherent experience. This work examines that capacity through the lens of entropy reduction and investigates whether quantum-inspired spiking neural architectures can illuminate the principles behind it. We analyze how biological neural networks compress, filter, and reorganize information through plasticity, inhibition, and sleep-dependent consolidation. These seemingly independent processes collectively drive the brain toward structured internal states that support stable cognition. In this project we develop an integrated quantum-inspired spiking neural network (Qi-SNN) that unifies these three distinct sleep functions: i) entropy reduction, ii) memory consolidation through structured replay iii) STDP plasticity. Building on a recent work in quantum state encoded-SNNs, we introduce a superposition based spike generator that linearly combines the complementary spike patterns through amplitude weight mixing, analogous to quantum superposition of basis states. During the wake state, the network learns with classical STDP on temporally encoded spike patterns. During sleep, we introduce a superposed replay mechanism, where amplitude-mixed spike trains are generated via quantum-inspired transformations. These replay spikes drive STDP updates that selectively strengthen synapses belonging to correlated memory traces while globally suppressing uncorrelated connections. Across successive simulated “nights”, the network exhibits progressive sharpening of weights corresponding to the memory traces, consistent with biological consolidation, while the overall synaptic weight distribution becomes more structured, resulting a measurable reduction in Shannon entropy. Additionally, a real-time quantum-STDP module in the system demonstrates that quantum-inspired dynamics can implement online potentiation/depression driven by probabilistic measurement outcomes. By connecting entropy regulation in biological neural networks with entropy dynamics in quantum learning networks, this work provides a conceptual and technical foundation for understanding how structured cognition can arise from probabilistic substrates.