Activation function (AF) plays a crucial role in the forward and backward processes of deep neural model learning and it is intriguing to suspect that specific problems (e.g. Image Classification) call for specific activation functions that improve performance and optimize the training methods.
Goal : This project will try to find out if building a CNN model with different activation function would significantly affect model accuracy.
Objectives :
- Collecting the common activation functions related to image classification CNN model from the literature and CNN practices.
- Applying the collected AFs to the CNN model that used to classify images from the CIFAR-10 database.
- Analysing the impact of AF on accuracy and performance and report the results.