Coefficients of the regression model that multiply each chemical species concentration to obtain the Bayesian posterior prediction for the best performing chemical network in the AB-BA task.

Positive numbers indicate species that are more likely to have high concentration in the S→C environments, while negative numbers belong to species that are more prevalent in the C→S environment. The magnitude of the weight relate to the significance of the chemical. The Bayesian interpretation is consistent with our explanation for the learning mechanism (see text).