Timing in simple conditioning and occasion setting: a neural network approach

Area of Science:

• Computational neuroscience
• Behavioral psychology
• Artificial intelligence

Background:

• Pavlovian conditioning involves learning associations between stimuli.
• Existing models struggle to fully capture the temporal dynamics and varied roles of stimuli in conditioning.
• Understanding these mechanisms is crucial for explaining complex learned behaviors.

Purpose of the Study:

• To present a novel neural network model of Pavlovian conditioning.
• To incorporate a timing mechanism that predicts the timing and duration of unconditioned stimulus (US) presentation.
• To allow stimuli to function as simple conditioned stimuli (CS), occasion setters, or both.

Main Methods:

• Developed a neural network architecture where stimuli evoke multiple traces of varying duration and amplitude.
• Implemented a competitive association mechanism, both direct and indirect (via hidden units), between CS traces and the US.
• Integrated a timing mechanism to predict the precise moment and duration of US delivery.

Main Results:

• The model successfully predicts the value, moment, and duration of reinforcement.
• Stimuli can dynamically assume roles of CS or occasion setter based on timing.
• Competition between CSs is both associative and temporal, predicting US presence, intensity, and temporal characteristics.

Conclusions:

• The model advances understanding of Pavlovian conditioning by integrating temporal and associative learning.
• It offers a more comprehensive explanation for simple, compound, and occasion setting conditioning phenomena.
• This framework provides a powerful tool for studying the neural basis of predictive learning and timing.