How to use 100-Days-Of-ML-Code: a day-by-day machine learning path

What is 100-Days-Of-ML-Code?

The repository is a diary rather than a course. It is based on the 100 Days of Machine Learning Coding idea proposed by Siraj Raval, and the author posts a log entry for each day of study. Topics include data preprocessing, simple and multiple linear regression, logistic regression, K nearest neighbours, support vector machines, naive bayes, decision trees and random forests.

Each entry is short. Some have a link to a code file, some have an infographic image, and many point to lectures or videos for deeper study. The README notes that the author later switched to posting infographics on alternate days because of less available time.

• Topics are arranged roughly from basic data preparation to introductory neural networks.
• Code links point to Markdown files in the Code folder of the repository.
• Datasets are linked from the README's datasets folder.
• The README describes the repository as a log, and its last push was 2023-12-29.

Get the repository onto your computer

The README gives one install command, which clones the repository with git. The project does not describe an installable package, so there is nothing to build or configure beyond getting the files. You need git installed to run the command. The README lists no other prerequisites.

• Open a terminal and run the clone command below.
• Move into the new folder and browse the Code, Info-graphs and datasets folders named in the README.

Follow the day-by-day path

Read the README from top to bottom and use the day headings as a syllabus. A sensible order is to start with Day 1 on data preprocessing, then the regression and classification days, then the decision tree and random forest days. Each heading names the topic, so you can jump to a specific algorithm.

For each day, read the short note first, then open the linked code file or infographic. The notes are personal and brief, so the infographic and the linked lecture or video usually carry more explanation than the text.

• Days 2 to 4 cover linear and logistic regression.
• Days 9 to 16 cover support vector machines, K nearest neighbours and naive bayes.
• Days 23 to 25 cover decision trees.
• Days 26 to 32 cover linear algebra and calculus using 3Blue1Brown playlists.
• Days 33 and 34 cover random forests.

Use the scikit-learn code and outside links

Several entries use the scikit-learn library. For example, Day 14 describes an SVM implementation on linearly related data using the SVC classifier, and Day 16 covers the kernel trick. Use these as pointers: open the linked code file, run it in your own Python environment, and compare it with the scikit-learn documentation.

The repository also links to lecture series. Caltech's CS 156 lectures by Professor Yaser Abu-Mostafa are mentioned from Day 19, the Bloomberg machine learning lectures from Day 15, the Coursera deep learning specialization from Day 17, and the 3Blue1Brown videos on linear algebra, calculus and neural networks from Day 26 onward.

• Start a code link in a notebook or Python file you control, so you can change values and rerun it.
• Keep a list of the outside links you finish, since the README does not track progress for you.
• Use the Beautiful Soup mention on Day 21 as a starting point if you want to collect data for practice.

Where the repository falls short

The README is a log written during the author's study, not a maintained textbook. Some entries are one line, some links may be out of date, and the README does not explain every formula or step. Because the last push was 2023-12-29, the repository will not add newer material.

If you need complete, current tutorials, pair this repository with a maintained course or book. The README's comparable projects are fast.ai, the Hands-On Machine Learning with Scikit-Learn book repository and the scikit-learn tutorials.