Prediction of Dynamic Preference by Using Temporal Dominance of Sensations Data

Abstract

Temporal dominance of sensations (TDS) method is a method to record multiple sensory changes over time during an experience such as eating food. We recorded sensory changes and preference changes while eating strawberries by using TDS method. Dynamic sensory and preference changes were trained on a reservoir computing network model, which estimated preferences with high accuracy for untrained sensory data. Reservoir computing has attracted attention as a fast machine learning method suitable for time-series pattern recognition, and its application to the TDS method for handling multidimensional time-series data was demonstrated.