Winter 2021-22 semester, a joint effort by Ido Amit and Dorin Shteyman.
All projects instructions are attached to this repository.
Here we were presented with the Virus Test Challenge Dataset (VTC) containing labeled information from patients suffering from all sorts of diseases. Our goal was to understand this dataset and classify wether a patient is affected by Corona-Virus.
General description of each assignment:
Project 1- We implemented examined our dataset and its' properties, made data imputation and cleaning, outlier detection, normalization and feature selection.
Project 2- In this assignment we implemented two algorithms from scratch: k-NN and Soft-SVM. After implementing those, we practiced basic hyperparameter tuning (model selection) using the three different classifiers: k-NN, ID3, and Soft-SVM. In addition we implemented our SGD algorithm and practiced feature mapping.
Project 3- In this final assignment, we tried to predict a continuous label using our dataset, the probability of a patient to have Corona-Virus. Our goal here was to regress that probability, i.e., the “virus score”, with the cheap features we used in the previous assignments. As can be seen in the project, we tried different approachs of optimization as well as different regression models and feature maps.