ml-from-scratch-cpp

Implementation of various supervised and unsupervised machine learning algorithm in C++

Setup instructions.

1. It is assumed that cmake and g++ is installed.
2. Run the following commands in the project directiory

mkdir build && cd build
cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=ON ..
make -j

3. We have now build the project in the build directory you will see various executables like linear_regression.
4. To run Linear Regression from build folder

./linear_regression ../data/LinearRegression/california_housing.csv 9 0

Here, ../data/LinearRegression/california_housing.csv is path to dataset and 9 is the column where target variable is located and 0 clarifies that the target columnn in numerical.
See data directory to see various demo datasets.

Features Implemented and Demonstration

CSV Parser

The load_csv() function parses a CSV file and stores the numberical data while elegantly formatting it to be displayed on the console using .print_dataset() function.

Splitting data

We can now use test_train_splie() to split testing and training data.
Before splitting we also randomly shuffle our data.

Scaling Data

We use Z Score normalization i.e. divide by standard deviation and subtract by mean.

Linear Regression

We can perform linear regression on a dataset, it uses normal equation to calculate the parameters theta.

Logistic Regression

We can perform logistic regression on a dataset, it uses gradient ascent and sigmoid function to find the parameters

Naive Bayes

We have also impleamented Gaussian Naive Bayes used for classification.

Performace Benchmark

Linear Regression

We used linear regression on housing_dataset.
Results in scikit-learn

Result in our custom library

Logistic Regression

We used logistic regression on breast cancer dataset
Results in scikit-learn

Results in custom library

Naive Bayes

We use naive bayes on breast cancer dataset

Comparing it to performance in sklearn.