The work demonstrates the image classification with both the hinge loss and the softmax loss with better explanation of the mathematics behind.
For the demonstration we are using the CIFAR-10 dataset to classify the images and we compare the accuaracy for thse 2 losses and also the work demonstrates the pros and cons of these 2 different losses.
To download the dataset to your local machine run the bash file on get_datasets.sh or you can download from the reference link in the notebook attached.
Hope this work gives a great intution about SVM(Support Vector Machines) considering different losses.
The work is implented only using basic libraries such as numpy but not using any other high level API’s.
