SSCC TD: a serial and simultaneous configural-cue compound stimuli representation for temporal difference learning

Area of Science:

• Computational neuroscience
• Machine learning
• Cognitive psychology

Background:

• Traditional Temporal Difference (TD) learning models struggle with complex stimulus configurations.
• Configural stimuli, where compounds create emergent properties (configural cues), are not well-explained by existing TD frameworks.

Purpose of the Study:

• To introduce a novel representational framework for TD learning.
• To enable the computation of configural stimuli and configural cues within the TD paradigm.
• To extend the explanatory and predictive power of TD models to phenomena involving compound stimuli and context.

Main Methods:

• Development of the Simultaneous and Serial Configural-cue Compound Stimuli Temporal Difference (SSCC TD) model.
• The framework allows for the computation of both simultaneous and serial stimulus compounds.
• Incorporation of experimental context as a potential component of stimulus compounds.

Main Results:

• The SSCC TD model can successfully compute configural stimuli and configural cues.
• The model accommodates both simultaneous and serial stimulus compounds, including contextual elements.
• The framework broadens the scope of phenomena explainable by TD learning.

Conclusions:

• The SSCC TD model significantly enhances the TD learning paradigm.
• It allows prediction of effects dependent on compound stimuli functioning as a whole, such as pattern learning and serial discriminations.
• Context-related effects in learning can now be better modeled using this advanced TD framework.