Uploaded March 2023 | Updated September 2026, 2 weeks ago

---

A generalization of the nearest neighbor classifier is the k nearest neighbors classifier (KNN for sort). K is the number of neighbors that will play a role in the classification.

I'll teach you how to implement the K Nearest Neighbors classifier in JavaScript and use it for drawing recognition.

The Course: youtube.com/playlist?list=PLB0Tybl0UNfYe9aJXfWw-Dw_4VnFrqRC4

⭐️HOMEWORK⭐️
Try to calculate a probability for the classification to be correct, based on the number of neighbors of the predicted class. Most libraries (tensorflow, scikit-learn) support this feature and it can be a useful thing to have. Other variants of nearest neighbor classifiers exist as well, like a version weighted by distance or one that considers a fixed size neighborhood. Think you can implement any of these? Share your code with me on Discord
discord.com/invite/gJFcF5XVn9
I’ll showcase my favorite implementations in a future video.

⭐️LINKS⭐️

📁 Data
github.com/gniziemazity/drawing-data

💻 Code
tps://github.com/gniziemazity/ml-course
✔️ Use P5 to follow along
✔️ P6 is the code after this lesson


⭐️TIMESTAMPS⭐️
00:00 Introduction
00:47 Intuition for the K Nearest Neighbors
01:50 Implementing the K Nearest Neighbors Classifier
06:32 Visualizing the K Nearest Neighbors on the Chart
09:44 Homework (Neighbors)