ML

Code The code can be found within this PDF Can also be found within this repo at static/images/wide_and_deep/wide_and_deep.pdf Overview This is a passion project I’ve been working on to deepen my understanding of recommender systems in my free time. While I’ve previously implemented simpler methods—such as matrix factorization or with basic content-based filtering, this project is my attempt to implement a more elegant solution inspired by this paper. Architecture Summary The model processes inputs related to user–item pairs and leverages two components: ...

Note to reader I do not include large code snippets as this was an assignment for my deep learning course at Northwestern and I don’t want future students copying it. It exists in a private repo of mine containing my school assignments. I will try to do my best to illustrate the work I did with fewer code examples Github Link Private url Overview This assignment challenged us to learn more about sequence modeling, and better understand the conceptual pros/cons between RNNs and LSTMs. We implemented RNNs, LSTMs, and a deep reinforcement learning model to play pong. I find that the sequence modeling and autoencoder information is more interesting in this project, so I didn’t include it here. ...

Note to reader I do not include large code snippets as this was an assignment for my deep learning course at Northwestern and I don’t want future students copying it. It exists in a private repo of mine containing my school assignments. I will try to do my best to illustrate the work I did with fewer code examples Github Link Private url Overview This assignment focused on Generative Adversarial Networks (GANs), adversarial examples, model collapse, and latent interpolation. We will touch on Adversarial examples and GANs ...

Github Link Private url Presentation Link https://youtu.be/gkKt3p52Kz8 Overview This was Michael Bertagna and my final project for our introduction to machine learning class. We designed a convolutional neural network (CNN) that takes in images of cell types, and attempts to classify them. We were able to achieve significant improvements compared to the KNN baseline used. Results KNN Baseline Results precision recall f1-score support accuracy 0.27 259 macro avg 0.14 0.15 0.14 259 weighted avg 0.24 0.27 0.23 259 CNN Results precision recall f1-score support accuracy 0.71 5143 macro avg 0.66 0.54 0.55 5143 weighted avg 0.71 0.71 0.70 5143 Presentation ...