Introduction to Machine Learning

(1) Purpose and content of the course

This course is designed to provide participants with an understanding of machine learning, an AI technology, through hands-on experience. It is assumed that participants have already taken their first programming course or are able to use Python at the same level. The course will use machine learning algorithms, mainly scikit-learn, a machine learning library, to learn regression, classification, clustering, etc. using sample data. Deep learning will not be covered in this course, but students will be able to acquire the prerequisite knowledge needed to learn deep learning through experience. The goal is to be able to understand what machine learning can do, without going deeply into the specific algorithms inside.

(2) Contents of each session

Session 1: “Basic fundamentals of machine learning and regression

Assuming that this is the first time for students to learn about machine learning, the lecture will first explain the basic concepts of machine learning, such as its position in AI technology, types of machine learning, and problems handled by machine learning. The lecture will also explain scikit-learn, a library used in both Introduction to Machine Learning and Introduction to Deep Learning. Finally, regression will be explained, and students will experience machine learning using actual data sets.

Session 2: “Classification

The second session will deal with classification problems. Specifically, we will discuss the types of classification problems and the methods used for them, such as logistic regression, support vector machines, and decision trees, and use scikit-learn data sets to understand how each model works and its characteristics. Lazy learning algorithms such as the k nearest neighbor method will also be covered.

Session 3: “Clustering

In the first and second sessions, we constructed machine learning models using data with known answers. In the third session, we will focus on cluster analysis, which is a method for finding hidden structures in data with unknown answers, and belongs to the unsupervised learning method. Terms and metrics related to clustering will be explained, and hierarchical clustering will also be covered. In the programming part, you will experience k-means clustering using scikit-learn on a real data set.

Session 4: “Preprocessing and Feature Engineering.”

In the third session, we learned about machine learning algorithms. However, by devising a data set to be trained by machine learning, it is possible to make predictions with higher accuracy. In the fourth session, you will learn how to handle data containing missing values, preprocessing such as data set partitioning, and feature engineering. The lecture will also explain how to compress data by dimensionality reduction using principal component analysis, etc., and give students hands-on experience with actual data sets.

Session 5: “Ensemble Learning

Students will learn about ensemble learning, which combines the training models learned so far to create a better classifier, as well as the terms “bagging” and “boosting” related to it. LightGBM, a framework for gradient boosting algorithms, will also be discussed. We will also review the previous lectures. As an appendix, we will also briefly describe the mechanism of reinforcement learning, focusing on Q-learning. Since reinforcement learning does not use libraries, there will be more explanations of codes and formulas than in other sessions, but we aim to simplify it as much as possible so that even beginners can understand the outline.