FEUP-IART

This repository hosts our project for the Artificial Intelligence (IART) course unit of the Masters in Informatics and Computer Engineering at FEUP.

Introduction

The aim of this project was to develop a binary classifier for identifying pulsars from a set of candidates.

Main difficulties include:

• balancing of the dataset (91% of candidates belong to the negative class, i.e. not pulsars);
• size of dataset, which includes 17898 learning samples: sizeable but small for deep learning purposes;

Dataset

HTRU2 is a data set which describes a sample of pulsar candidates collected during the High Time Resolution Universe Survey (South).

Available here.

Tools

• Python3
• Keras
• Tensorflow (backend)
• Scikit-Learn
• TensorBoard

Install Dependencies

pip3 install -r requirements.txt

Contributors

• André Cruz
• Edgar Carneiro
• João Carvalho