Predicting Straight-Edge Diffraction using a Neural Network

Abstract

In this experiment we studied the physical effect of diffraction through a straight edge and used a neural network to model and predict the behavior given its parameters. Straight edge diffraction studies how light will interfere with itself when directed towards a straight opaque edge. We used the apparatus to generate a set of data to train on a neural network to predict the diffraction pattern. By using Fresnel’s mathematical model of predicting a diffraction pattern, we trained the neural network with the noiseless theoretical data. The neural network was then able to predict the sliced diffraction pattern to an accuracy of 99.5%.