General overview: In this homework you will learn how to implement and test neural network models for solving unsupervised problems. For simplicity and to allow continuity with the kind of data you have seen before, the homework will be based on images of FashionMNIST. However, you can optionally explore different image collections (e.g., Caltech or Cifar) or other datasets based on your interests. The basic tasks for the homework will require to test and analyze the convolutional autoencoder implemented during the Lab practice. If you prefer, you can opt for a fully-connected autoencoder, which should achieve similar performance considering the relatively small size of the FashionMNIST images. As for the previous homework, you should explore the use of advanced optimizers and regularization methods. Learning hyperparameters should be tuned using appropriate search procedures, and final accuracy should be evaluated using a cross-validation setup. More advanced tasks will require the exploration of denoising and variational / adversarial architectures.
Technical notes: The homework should be implemented in Python using the PyTorch framework. The student can explore additional libraries and tools to implement the models; however, please make sure you understand the code you are writing because during the exam you might receive specific questions related to your implementation. The entire source code required to run the homework must be uploaded as a compressed archive in a Moodle section dedicated to the homework. If your code will be entirely included in a single Python notebook, just upload the notebook file.
