Training Data and the Maladaptive Mind

Abstract

Childhood trauma reframed through machine learning training data quality: extreme penalties cause gradient cascades (1,247× amplification, p<0.001), noisy signals produce behavioral instability, absent positive examples create emotional recognition deficits (alexithymia). PyTorch experiments validate computational mechanisms.