Inverse Problem Approach to Machine Learning with Application in the Option Price Correction

Abstract

We investigate a new method in learning to fix the existence of an unsuitable subfunction of a general system. We assume this subfunction is dependent on the system input variables. In this process, we put a learner instead of the unsuitable subfunction and train it by a training model obtained from inverse problems and fractional derivatives, respectively. Finally, we implemented this method on a simple financial model and examined the results with simulated and real data.