Lesson 3 Project
Classification Neural Networks
Complete this project to enhance the classification network from the code video - and then to make a new type of classification network.
Use the classification network from the video, but change it to use a PyTorch Dataset and DataLoader
HINT:Look back and lesson 2 for more information about how to make a Dataset and DataLoader.
After your network uses a Dataloader, split the data into training and validation sets - and then compute the accuracy on the validation set as well as the training set and compare them.
Use the accuracy values to make sure that the network isn't overfitting or underfitting.
Use the accuracy values to make sure that the network isn't overfitting or underfitting.
HINT:If the network is overfitting, look back at lesson 2 to get ideas about how to fix that.
Find a dataset that only has 2 classes, and use binary cross entropy to predict the correct class. If you can't find a dataset, Here is a dataset from Kaggle that you could use: https://www.kaggle.com/mlg-ulb/creditcardfraud
Find or make a multi-label dataset (where each item can belong to 0 or more classes) and create a network to do class prediction
HINT:Remember that you can use binary cross entropy across multiple inputs to get independent class perditions
