miniml-adv 0.1.0

Machine Learning Library Built From Scratch Using NumPy

MiniML

A Machine Learning Library Built From Scratch Using NumPy.

MiniML is a machine learning library implemented completely from scratch using NumPy without relying on scikit-learn for core algorithms.

The goal of this project is to deeply understand:

• Machine Learning Algorithms
• Optimization
• Neural Networks
• Backpropagation
• Linear Algebra
• ML Engineering
• Software Architecture

The project includes:

• Classical ML Algorithms
• Clustering Algorithms
• Dimensionality Reduction
• Neural Network Engine
• Optimizers
• Visualization Tools
• Benchmarking System
• CLI Interface

Features

• Linear Regression
• Logistic Regression
• KNN
• KMeans
• DBSCAN
• PCA
• Neural Networks
• SGD Optimizer
• Momentum Optimizer
• Adam Optimizer
• Mini-Batch Training
• Visualization Tools
• Benchmarking
• CLI Tool

Installation

git clone <repo-link>

cd MiniML

pip install -r requirements.txt

Quick Start

from miniml.linear_regression import LinearRegression

model = LinearRegression()

model.fit(X, y)

predictions = model.predict(X)

Neural Network Example

from miniml.neural_network import NeuralNetwork

from miniml.layers import Dense
from miniml.activations import ReLU, Sigmoid

from miniml.optimizers import Adam

model = NeuralNetwork()

model.add(Dense(2, 8))
model.add(ReLU())

model.add(Dense(8, 1))
model.add(Sigmoid())

model.compile(
    loss="binary_cross_entropy",
    optimizer=Adam(learning_rate=0.01)
)

model.fit(X, y, epochs=1000)

Project Architecture

MiniML follows modular ML framework design:

• Layers
• Activations
• Loss Functions
• Optimizers
• Training Engine
• Visualization System
• Benchmarking System

Neural networks are implemented using:

• Forward Propagation
• Backpropagation
• Gradient Descent
• Automatic Gradient Flow

Mathematical Concepts Implemented

• Gradient Descent
• Binary Cross Entropy
• Backpropagation
• Covariance Matrices
• Eigen Decomposition
• Distance Metrics
• Clustering Optimization
• Numerical Stability
• Vectorized Computation

Benchmarking

MiniML includes benchmarking tools comparing:

• Training Time
• Prediction Time
• Accuracy
• Memory Usage

against scikit-learn implementations.

Visualization Tools

• Loss Curves
• Accuracy Curves
• Decision Boundaries
• Cluster Visualizations
• PCA Projections

CLI Usage

python cli.py --model linear_regression

python cli.py --model logistic_regression

python cli.py --model kmeans

python cli.py --model neural_network

Folder Structure

MiniML/
│
├── miniml/
├── benchmarks/
├── visualizations/
├── examples/
├── tests/
├── datasets/
├── cli.py
├── README.md
└── requirements.txt

Future Improvements

• CNN Layers
• Transformer Architecture
• GPU Support
• Automatic Differentiation Engine
• Model Serialization
• Hyperparameter Tuning
• Distributed Training
• CUDA Acceleration

---