An end-to-end deep learning pipeline for automatic sleep stage classification from polysomnography (PSG) signals. The system classifies 30-second EEG/EOG/EMG epochs into 5 AASM sleep stages (Wake, N1, N2, N3, REM) using a dual-input Teacher model (CNN + Transformer, κ=0.636) distilled into a lightweight Student model

In a zero-logic ecosystem, 512 autonomous agents are tasked with navigating a environment governed strictly by thermodynamic constraints rather than predefined heuristic rules. To endure a state of chronic computational energy scarcity, these agents must evolve persistent "habits"—automated, low-entropy behavioral patterns that minimize energy loss