In this project we have used Genetic Algorithm to do Neural Architecture Search(NAS) for finding the best performing Convolution Neural Network(CNN) architecture on the fashion-mnist dataset.
Our objective was to find a model that gives highest accuracy while using least number of parameters. Thus we used the fitness_score = accuracy/log(numberof parameters).
We used the following genetic algorihtm startegy to achive this:
Following is the scatter plot of various models, their accuracy and fitness scores.
Please check the attached report for more details.