My name is Jeremy Di Dio and I am currently pursuing a master degree in Data Science at EPFL.

Here are some projects I worked on:

• Real-Time Emotion Recognition in VR

  In this semester project done at the Immersive Interactive Group at EPFL. The goal was to implement a machine learning model able to detect in real-time the emotions (Anger, Happiness, Disgust, Sadness, Fear and Surprise) from a particular type of data provided by a new facial tracker used in virtual reality environment. This Streamlit app present the evaluation part of the implemented models.

  Here is the final report of this project.

• Tweets sentimental analysis

  In this project, we developed multiple machine learning models for the purpose of conducting sentiment analysis on Twitter posts. The objective was to classify the posts as either positive or negative. To achieve this, we employed some state-of-the-art transformer models, including BERT and RoBERTa, for the classification task. The performance of our best model was evaluated on the testing set, yielding an accuracy of 0.892.

• QAM modulation for complex AWGN communication

  Course project for COM-302 (Principles of Digital Communication). Done in group of 4, the goal of this project was to use QAM modulation and demodulation to correctly communicate through a complex AWGN channel.

  
  

• Higgs boson machine learning challenge

  In this project, we present our methodology to answer the Higgs boson machine learning challenge which consists in building a binary classifier to predict whether an event corresponds to the decay of a Higgs boson or not. In other words, we will train machine learning model on precollected data to further predict wether a collision was signal or background.