🧠 ML Systems Lab

A hands-on repository for implementing machine learning algorithms from scratch, with a focus on understanding the mathematics, optimization, and system design behind modern ML models.

This project bridges the gap between theory and real-world ML engineering.

---

🚀 Overview

Most machine learning libraries hide the underlying mechanics.
This repository focuses on:

• Building algorithms from first principles
• Understanding how models learn
• Writing clean, modular, production-style Python code
• Developing intuition for model performance and limitations

---

🎯 Objectives

• Implement core ML algorithms without high-level frameworks
• Strengthen understanding of optimization and learning dynamics
• Build reusable ML components
• Prepare for advanced ML systems and MLOps workflows

---

📂 Project Structure

ml-systems-lab/
│
├── src/
│ ├── perceptron/
│ │ ├── perceptron.py
│ │ ├── xor.py
│ │ └── init.py
│ ├── backpropagation/
│ │ ├── backpropagation.py
│ │ └── init.py
│ │
│ ├── linear_models/
│ │ ├── linear_regression.py
│ │ ├── logistic_regression.py
│ │ └── init.py
│ │
│ ├── utils/
│ │ ├── metrics.py
│ │ ├── data_utils.py
│ │ └── init.py
│ │
│ └── init.py
│
├── notebooks/
│ └── experiments.ipynb
│
├── tests/
│ ├── test_perceptron.py
│ │ ├── xor.py
│ └── test_linear_models.py
│
├── data/
│ └── (datasets go here)
│
├── examples/
│ └── perceptron_example.py
│
├── requirements.txt
├── README.md
└── .gitignore

---

⚙️ Implemented Algorithms

✅ Supervised Learning

• Perceptron (binary classification)
• XOR (basic multilayer perceptron)

🔜 Coming Soon

• Linear Regression
• Logistic Regression
• Neural Networks (MLP)
• Decision Trees
• Support Vector Machines (SVM)

---

🧩 Example Usage

import numpy as np
from src.perceptron.perceptron import Perceptron

X = np.array([[0,0], [0,1], [1,0], [1,1]])
y = np.array([0, 0, 0, 1])

model = Perceptron(input_dim=2, learning_rate=0.1)
model.train(X, y)

---

🛠️ Key Features

• Pure NumPy implementations (no ML frameworks)
• Modular and extensible design
• Clear training loops and update rules
• Easy to debug and experiment with
• Structured like a real ML codebase

📊 Design Philosophy

• Clarity over cleverness
• Understanding over abstraction
• Engineering discipline over shortcuts

Each algorithm is implemented with:

• Direct math-to-code mapping
• Explicit training procedures
• Reusable components

---

🧑‍💻 Tech Stack

• Python
• NumPy
• (Planned) Pandas
• (Planned) Matplotlib

---

💡 Author

Derrick Nyongesa

Data Scientist | Electrical & Electronics Engineer

Focus: Understanding Machine Learning Algorithims