Uploaded August 2023 | Updated September 2026, 2 weeks ago

---

In this lesson we use Python's Scikit-learn library to optimize a deep neural network where the input layer contains individual pixel intensities (a feature vector with 400 dimensions). We then use this optimized neural network in JavaScript to achieve an accuracy above 80%.

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

Phase 3 Poll:
forms.office.com/e/QTMCLLaV24

💻 Code
github.com/gniziemazity/ml-course-phase-2
✔️ Use P8 to follow along
✔️ P9 is the code after this lesson

Scikit-learn documentation:
scikit-learn.org/stable/modules/generated/sklearn.neural_network.MLPClassifier.html

⭐️TIMESTAMPS⭐️

00:00 Course Introduction
01:08 Using Pixels as Features
07:02 Improved Visualizer
08:49 Experimenting with Different Neural Network Structures
11:25 Conclusions