Rescorla-Wagner Blocking and Overshadowing: Prediction-Error Learning of Cue Associations
Train a learner with a bell and a lamp before food and watch associative strengths change by prediction error, as in the Rescorla-Wagner model of classical conditioning. Pre-train the bell alone and the lamp is blocked; present both from the start and they overshadow each other in proportion to their salience. Set salience, learning rate, food size and the number of trials in each phase, with a Blender-built conditioning chamber and independent numerical checks of the closed-form equations.
Train a learner with a bell and a lamp before food and watch associative strengths change by prediction error, as in the Rescorla-Wagner model of classical conditioning. Pre-train the bell alone and the lamp is blocked; present both from the start and they overshadow each other in proportion to their salience. Set salience, learning rate, food size and the number of trials in each phase, with a Blender-built conditioning chamber and independent numerical checks of the closed-form equations.
Train a learner with a bell and a lamp before food and watch associative strengths change by prediction error, as in the Rescorla-Wagner model of classical conditioning. Pre-train the bell alone and the lamp is blocked; present both from the start and they overshadow each other in proportion to their salience. Set salience, learning rate, food size and the number of trials in each phase, with a Blender-built conditioning chamber and independent numerical checks of the closed-form equations.
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install