KIOS Summer School 2018: Lab & Hackathon
As part of the summer school, a 3-hour laboratory session and a 2-day Hackathon Competition on Machine Learning (ML) will be organized.
The Laboratory will consist of two 1.5-hour sessions explaining and demonstrating the art of solving supervised machine learning problems using state-of-the-art software tools and ML libraries (e.g., Python, Pandas, Jupyter Notebook, Scikit-Learn). The material for the lab, along with detailed instructions on how to use it, can be found here: https://github.com/KIOS-Research/KIOS-Graduate-Summer-School-2018
The first session will concentrate on classification and regression using various learning machines (e.g. linear/logistic regression, decision trees, support vector machines, neural networks, random forests, etc.) and multiple datasets. In addition to experimenting with different learning machines, the trainees will also have the opportunity to acquire valuable knowledge on practical aspects of supervised learning to maximize performance such as data pre-processing (e.g., standardization, imputation of missing values), dimensionality reduction, over-fitting avoidance, as well as model selection and evaluation.
The second session will focus on system identification and time-series anomaly detection using ML tools. System identification will be examined both for linear and nonlinear system models using kernel-based regularized least squares methods. The importance of regularization on identifying the system rank and avoiding outliers will be demonstrated. Anomaly detection will be examined using various novelty/outlier detection techniques such as the one-class support vector machines and the local outlier factor algorithms.
The hackathon competition will run for two days of the Summer School and feature a number of challenging real-life supervised ML problems (similar to the ones examined during the laboratory sessions). The trainees will work in teams of 3-4 persons to solve the problems and submit their answers to an online system for evaluation. The teams with the highest scores will receive prizes and certificates.