• Reports of 1 and 2

• In this homework, you are invited to provide solutions for two 10-class classification problems with medium and large input space.
• The two datasets are available in this folder
• To solve the problem, you can use any method, any tool, any programming language. Exception: you cannot use solutions based on neural networks, as this will be the topic of Homework 2. It is your choice to pre-process the input data in any way it is useful for the method you are using.

• In this homework, you are invited to solve an image classification problem to learn the behaviour of a racing car in a Gym environment.
• The dataset (divided into training and test sets) contains 96x96 color images labelled with one out of the 5 actions available for the control of the car.
• The dataset is provided as sets of images organized in folders labelled with the id of the action.
• The datasets are available in this folder

• define your own model based on CNN (do not use pre-defined architectures or pre-trained models)
• solve the problem using 2 different approaches. By approaches, we mean architectures, optimizers, regularizations, preprocessing, etc., not just different values of the hyperparameters.
• report analysis of fundamental metrics, such as accuracy, f1, precision, and recall. Screenshoots of the output of some model is not sufficient. You must elaborate the results providing better visualizations and specific comments.
• conduct an analysis of at least one hyperparameter in relation to the chosen metrics.

In practice, you have to define two different approaches, train them with different hyperparameters, compare them (performance vs hyperparameter), visualize the results in proper forms (figures, tables), and discuss the results. Any software library can be used, including Python (tensorflow or pytorch), MATLAB, etc.