- 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.shor 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.