#machine-learning-labs Lab Questions for 2301793-2019 class at Chulalongkorn University (Special Topics in CS - Modern Machine Learning)
This repo contains 8 labs for the class in the form of Jupyter notebooks. You can either run in your local machine or in Google Colab.
To import the code to Google Colab, please follow this guide (you can skip step 2). Most of the libraries should have already been pre-installed in Colab.
Getting started with popular machine learning libraries and concepts. For this course, we'll mainly use scikit-learn and PyTorch.
In this lab, you'll get the flavour of the classification task using a toy dataset from Tensorflow playground. We'll also explore building a simple logistic regression model using Tensorflow.
We'll start the lab with a classic donut problem, which is one of the simplest problem that's not linearly separable. This gives us a chance to try out different SVM kernels. After that, we'll tap into the realm of multi-class classification with a text dataset crawled from Wikipedia.
We'll explore a popular toy dataset Fashion-MNIST, which is a slightly more difficult task than the classic MNIST. You're classify images into ten cloth categories and perform some standard analysis.
In this lab, you'll apply random forest algorithm to classify a real-world housing price dataset. Besides, you'll also use the built-in feature importance method to perform feature selection.
In this lab, you'll try cluster points using sklearn's k-means algorithm as well as have a chance to implement an online version of the algorithm from scratch.
In this lab, you will learn how to apply a gradient boosted model on a classic Kaggle competition - Titanic: Machine Learning from Disaster. You'll use the famous library xgboost for this task.
In this lab, you'll explore various popular multi-armed bandit algorithms for exploration and exploitation.
Some of the materials might be taken from other people's code, including but not limited to the following list: