IntelliNeuro 0.1.0

Perceptron implementation for linear, binary, and multi-class tasks

README: |

🚀 IntelliNeuro PerceptronX

🤖 Custom Perceptron Machine Learning Library

Author: Ajay Soni

---

📚 Introduction

IntelliNeuro’s PerceptronX is a fully custom-built Perceptron machine learning library created from scratch, utilizing only NumPy and pandas for efficient numerical computations and data handling.

This library provides:

• 🔹 Linear regression
• 🔹 Binary classification with sigmoid activation
• 🔹 Foundational multi-class classification using softmax activation

It is designed with educational clarity and modularity in mind, perfect for learners and practitioners eager to grasp the inner workings of perceptrons through clean, step-by-step gradient descent optimization.

---

✨ Key Features

• 🎯 Versatile Learning Tasks

  • Linear regression for continuous value prediction
  • Binary classification with sigmoid activation
  • Multi-class classification via softmax (demo-level support)

• ⚙️ Robust Preprocessing Support

  • Built-in scaling options: none, minmax, and standard
  • Validation to ensure correct scaling usage

• ⚡ Efficient Training Pipeline

  • Gradient descent optimization with up to 2,500,000 iterations
  • Early stopping based on tolerance threshold
  • Customizable learning rate and validation split
  • Detailed verbose training logs for monitoring

• 📊 Comprehensive Prediction & Evaluation

  • Seamless predictions for all supported tasks
  • Wide range of evaluation metrics:
    • Regression: MSE, RMSE, RMSLE
    • Binary classification: Accuracy, Precision, Recall, F1
    • Multi-class classification: Accuracy, Weighted Precision, Recall, F1

• 👌 User-Friendly & Informative

  • Clear, color-coded terminal outputs (via colorama)
  • Helpful warnings and error messages to guide usage

---

🚀 Quickstart Guide

1. Install and Import the library:

   pip install IntelliNeuro==0.1.0
   from PerceptronX import Perceptron
   

2. Initialize the model:

   model = Perceptron(
       learning_rate=0.001,
       validation_split=0.2,
       scaling='none',  # Options: 'none', 'minmax', 'standard'
       is_scaled=False,
       tolerance=1e-6
   )
   

3. Train your model:

   model.fit(X_train, y_train)
   

4. Make predictions:

   predictions = model.predict(X_test)
   

5. Evaluate model performance:

   score = model.score(X_test, y_test, metrics='accuracy')
   print(f"Model Accuracy: {score}")
   

---

🔍 How It Works

• ⚙️ Gradient Descent:

  • Iteratively updates weights and bias to minimize loss
  • Different loss functions per task:
    • Linear regression: Mean Squared Error (MSE)
    • Binary classification: Log loss with sigmoid activation
    • Multi-class classification: Cross-entropy with softmax

• 🔄 Scaling Techniques:

  • Manual min-max and standard scaling implementations
  • Auto-validation to prevent misuse

• 🧪 Validation:

  • Splits data based on validation_split parameter
  • Prints validation metrics post-training for model monitoring

• 🔔 Activation Functions:

  • Sigmoid for binary classification tasks
  • Softmax for multi-class classification

---

⚠️ Important Notes

• Multi-class classification is experimental and intended for learning purposes only.
• Ensure your data is scaled appropriately or specify scaling parameters.
• Predictions or scoring before training will raise errors.
• Color-coded outputs help distinguish warnings, info, and errors clearly.

---

🛠️ Installation

Requires Python 3.7+ and the following packages:

• numpy
• pandas
• scikit-learn (for evaluation metrics)
• colorama (for colored terminal output)

Install dependencies with:

pip install numpy pandas scikit-learn colorama

---

💡 Best Practices

• Always preprocess your features correctly or enable built-in scaling.
• Use validation_split to track model performance during training.
• Treat multi-class functionality as a conceptual demonstration.
• Tune learning_rate and tolerance to balance accuracy and training speed.
• Watch console warnings carefully to avoid common pitfalls.

---

🤝 Support & Contributions

Author: Ajay Soni
Repository: https://github.com/ml-beginner-learner/IntelliNeuro
Email: [e-mail]

Contributions, issues, and feedback are welcome!
Please open an issue or pull request to help improve this library.

---

📄 License

This project is licensed under the MIT License.
Feel free to use, modify, and distribute with proper attribution.