Assignment 3 – ML Course

Students:
Shir Nehoray 209544352
Itamar Talpaz 315078196

GitHub repository:
https://github.com/talpazi/ml-course-assign3

---

Repository Description

This repository contains the solution for Assignment 3, which extends the original Chapter 11 implementation from Machine Learning with PyTorch and Scikit-Learn (Raschka et al., 2022) by adding a second hidden layer and evaluating the model on the MNIST dataset.

The project includes three implementations:

1. Original single-hidden-layer ANN (baseline from the book)
2. Revised two-hidden-layer ANN implemented from scratch using NumPy (with softmax output)
3. Equivalent two-hidden-layer ANN implemented using Keras (TensorFlow)

---

Repository Structure

ml-course-assign3\
├── assign3-notebook.ipynb # Main notebook containing all implementations, and analysis\
├── README.md # Project description and repository structure\
├── python_environment_check.py# Package version check script (from Raschka et al., 2022)\
├── saved_models/ # Saved model weights, results, and training histories\
├── summry_plot.png # Summary comparison plots\
├── summry_plot2.png # Additional training and evaluation plots

Notes

• Detailed explanations and results are provided in the accompanying PDF report submitted via Moodle.