BasicML

Learning basics of Machine learning

Language : Python - numpy, matplotlib, seaborn, scikit-learn, pytorch

This repository is a comprehensive portfolio documenting my journey into Machine Learning, from foundational theory to practical application. It showcases a commitment to understanding algorithms from first principles and applying them to solve real-world problems.

The core of this repository is MahaML, a custom Python package where I have implemented major machine learning algorithms from scratch using only NumPy. This project was undertaken to build a deep, mathematical understanding of the models that power modern data science.

Setup and Installation

Step 1: Clone the Repository

git clone https://github.com/Mahanth-Maha/BasicML.git
cd BasicML

Step 2: Create and Activate Environment

Using Conda:

conda create --name basicML python==3.9 -y
conda activate basicML
conda install -c anaconda ipykernel -y
python -m ipykernel install --user --name basicML --display-name "BasicML"

Install dependencies:

pip install -r requirements.txt

Step 3: Launch Jupyter

jupyter lab

Optional – Display All Output Cells

In a notebook:

from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_interactivity = "all"

Folder Overview

MahaML/

A self-written ML algorithms package using only numpy and matplotlib. Each algorithm is implemented from first principles for a complete mathematical understanding.

Includes:

• Regression and classification models
• Ensemble methods (Bagging, Boosting, Stacking)
• Clustering and dimensionality reduction prototypes
• Helpers and testing utilities

Example:

python MahaML/linearRegression.py

ML_NLP_basics/

Contains notebooks demonstrating:

• Exploratory Data Analysis (EDA)
• Text preprocessing
• Supervised and unsupervised ML workflows
• Building algorithms step-by-step and testing on datasets

Datasets include: IMDb reviews, spam mails, student grades, housing data, and more.

python_basics/

Notebooks covering:

• Python data types and structures
• User-defined functions and modules
• Plotting and visualization
• Probability and linear algebra essentials

pytorch_basics/ and DL_basics/

Early work in deep learning. Focus on understanding tensor operations, autograd, and simple feedforward architectures.

kaggle/

Contains scripts, data, and submissions from Kaggle competitions such as:

• Titanic Survival Prediction
• IMDB Genre Classification
• Playground Series Experiments

Projects/ (Keeping them in local - in progress)

Applied ML and DL projects demonstrating practical problem-solving:

• CTR (Click-Through-Rate) Prediction
• News Recommendation System (EBNeRD dataset)
• Online Shoppers Purchase Intention Analysis

Courses/

Lecture slides and references from Neural Networks for Machine Learning and other learning materials.

Notes/

Self-written markdown summaries and formula sheets for quick reference.

MahaML Library: Implementation Progress

This section tracks the development status of the from-scratch algorithm implementations within the MahaML package.

Supervised Learning

• Regression

  • Linear Regression
  • k-Nearest Neighbors (k-NN)

• Classification

  • Logistic Regression
  • k-Nearest Neighbors (k-NN)
  • Naive Bayes
  • Support Vector Machine (SVM)
  • Decision Tree

• Ensemble Methods

  • Bagging (General + Random Forest)
  • Boosting
  • Stacking

Unsupervised Learning

• Clustering
  • k-Means Clustering (in progress)

Future Development Roadmap

• Clustering - DBScan, Hierarchical
• Dimensionality Reduction (e.g., PCA, T-sne)
• Recommender System Algorithms

Example Workflow

1. Start by reviewing foundational notebooks in python_basics/.
2. Explore algorithm implementations in MahaML/ — for instance, linearRegression.py.
3. Compare your implementation’s output with scikit-learn equivalents in ML_NLP_basics/.
4. Progress into neural networks using notebooks under pytorch_basics/.
5. Apply knowledge to real datasets in the kaggle/ and Projects/ directories.

Technologies Used

• Python 3.9+
• NumPy, Matplotlib, Seaborn
• scikit-learn
• PyTorch
• NLTK, pandas
• Jupyter Notebook / VSCode

License

This repository is open for educational and learning purposes. Feel free to fork, explore, or contribute improvements with proper attribution.

My Useful Setup codes (not related to this repo)

OS : Windows (Cmd/PS)

Setup a venv and create a kernal for vscode

conda create --name <NAME> python==<PyVersion> -y
conda activate <NAME>
conda install -c anaconda ipykernel -y
python -m ipykernel install --user --name <NAME> --display-name <"Kernal_Name">

pip install requirements.txt

• change <NAME> with env-name (EX: torchEnv)
• change <"Kernal_Name"> with env-name (EX: "trenv" , note:in string quotes)

To set jupyter notebook to print all / last expressions

from IPython.core.interactiveshell import InteractiveShell
InteractiveShell.ast_node_interactivity = "last"  # Options: 'last', 'all', 'last_expr'

Contact

Author: $\large \textrm{Mahanth Yalla (Maha)}$, MTech AI, IISc Bangalore

Focus Areas: Machine Learning, Deep Learning - CV & NLP

Portfolio: mahanthyalla.in

GitHub: github.com/Mahanth-Maha