ML-CPP-Core

ML-CPP-Core is a lightweight C++ library implementing fundamental machine learning algorithms from scratch: Linear Regression, Decision Trees, K-Means Clustering, and a Feedforward Neural Network. This repository is designed for educational purposes, offering clean, minimal implementations to help you understand the core mechanics of these algorithms. Ideal for students, hobbyists, and developers learning machine learning concepts in C++.

Features

Linear Regression: Fits a line to data using the least squares method for regression tasks.
Decision Tree: Implements a binary classification tree with Gini impurity-based splitting.
K-Means Clustering: Groups data points into k clusters based on Euclidean distance.
Neural Network: Feedforward neural network with one hidden layer, trained via backpropagation (demonstrates XOR solution).
Example Usage: The main program demonstrates each algorithm with practical examples.

Getting Started

Prerequisites

C++ compiler supporting C++11 or later (e.g., g++, clang++)
Standard Template Library (STL)
No external dependencies required

Installation

git clone https://github.com/murugiclin/ML-CPP-Core.git
cd ML-CPP-Core
g++ MachineLearning.cpp -o ml_core
./ml_core

Usage

The main() function in MachineLearning.cpp demonstrates example usage for each algorithm:

Linear Regression: Fits a line to sample data (e.g., y = 2x) and predicts a new value.
Decision Tree: Classifies 2D points (e.g., binary classification on a simple dataset).
K-Means: Clusters 2D points into two groups.
Neural Network: Trains on the XOR problem to demonstrate non-linear classification.

To experiment, modify the main() function with your own datasets or parameters.

Code Structure

LinearRegression: Class implementing linear regression with fit and predict methods.
DecisionTree: Class for binary classification with configurable max depth.
KMeans: Class for clustering with configurable cluster count.
NeuralNetwork: Feedforward neural network with one hidden layer, trained via backpropagation.
All implementations: Contained in MachineLearning.cpp.

Example Output

Linear Regression: y = 2x + 0
Prediction for x=6: 12
Decision Tree prediction for [0.5, 0.5]: 0
K-Means cluster assignments: 0 0 1 1 1
Epoch 0 Error: 0.25
...
Predictions:
Input: [0, 0] Output: 0.03
Input: [0, 1] Output: 0.95
Input: [1, 0] Output: 0.94
Input: [1, 1] Output: 0.06

Notes

These are simplified implementations for educational purposes. They lack optimizations like batch processing, advanced regularization, or vectorized operations.
The Decision Tree uses a basic Gini impurity criterion and does not handle continuous outputs.
K-Means uses random centroid initialization, which may lead to varying results.
The Neural Network is a basic feedforward model with sigmoid activation and mean squared error loss.
For production use, consider established libraries like MLpack, Eigen, or TensorFlow C++ API.

Contributing

Contributions are welcome! Please submit a pull request or open an issue for bug reports, suggestions, or improvements. Potential areas for enhancement:

Adding more algorithms (e.g., SVM, KNN)
Optimizing performance with vectorization
Supporting additional activation functions or loss metrics

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Inspired by the need for clear, educational machine learning implementations in C++.
Thanks to the open-source community for resources on algorithm fundamentals.

NeuralNetwork

Install / Use

README