Minimal Neural Architecture for Conflict-Based Behavior Generation and Memory Evolution 2.0

📖 Описание: This dataset represents a minimal neural agent architecture — NeuroATM_2.1 — that demonstrates adaptive behavior through internal conflict resolution. The system detects when no known response applies, triggers a new behavior (called a "third path"), and stores it for future re-use. Each simulation run produces a unique memory graph showing how behavior emerges and stabilizes over time. Unlike standard machine learning models, NeuroATM does not rely on large datasets or predefined reward schemas. Instead, it evolves from contradictions and internal logic, making it a candidate framework for exploring minimal artificial cognition. The package includes reproducible notebooks, screenshots from multiple runs, and a short summary of the system's architecture and implications. Author: Bato Naidanov Автор: Бато Найданов ([email protected]) № 1-14916778941 Телеграмма: @BATONAIDANOV