SAMVEDA: An Experience-Driven Cognitive Architecture for Predictive Memory and a Framework for Continual Learning

SAMVEDA is an experience-driven cognitive architecture designed around event-based memory, sequential prediction, prediction error, and adaptive knowledge formation. The architecture represents experience as interconnected events and learns transition structure from repeated observations. The implemented SAMVEDA Core demonstrates event observation, transition learning, next-event prediction, prediction confidence, prediction error, spreading activation, recall, and memory-graph inspection. A documented SAMVEDA Core v4.0 experiment demonstrates sequential pattern learning and next-event prediction. The work further proposes a continual-learning extension in which prediction error and evidence accumulation drive change detection, adaptive updating, memory consolidation, and recovery of previously learned knowledge. A controlled A–B–A benchmark with explicit baselines is provided as an experimental framework for evaluating adaptation and retention. The work is presented as an initial cognitive-architecture framework and does not claim human-like cognition or general intelligence. The continual-learning and broader cognitive claims remain subjects for further empirical evaluation.