Using lower note ratios to create a reward function for music reinforcement learning algorithm

Question

I am creating a music learning reinforcement algorithm. Would a good reward function reward highly movements that result in ratios with the lowest integer numbers used within a chord and between the superset of note ratios between the notes in consecutive chords in a progression? Also What is the relationship between two chords whose superset of ratios between notes is the lowest. Is it the same as adding the frequencies in each chord and finding the ratio? Also could i make this comparison more long ranging by generalizing it, such that the sum of the frequencies of notes in 4 bars should have the lowest ratio with the sum of frequencies in the next 4 bars.? Depending on the previous answer?Ofcourse i could simply calculate this and move my chords in this direction, but i have another source of information. So all things being equal the notes should harmonize with the lowest integer values on the ratios.