---

   ||    // \\ | || | ||    ||\ || | || |
   ||    ||=||   ||   ||==  ||\\||   ||  
   ||__| || ||   ||   ||___ || \||   ||    
   ____    ___     ___    __  ______
      || ))  // \\   // \\  (( \ | || |
      ||=)  ((   )) ((   ))  \\    ||  
      ||_))  \\_//   \\_//  \_))   ||
        __                         
       /   |  _   _  _ . (_ .  _  _ 
       \__ | (_| _) _) | |  | (- |
         __       ___           
        /_ |     |__ \      |    | 
         | |        ) |     |    | 
         | |       / /      |    |  
         | |  __  / /_   __ |____|__
         |_| (__)|____| (__)     |

---

---

LatentBoostClassifier: Hybrid Generative Model: CVAE + CGAN + RF

---

---

| http://dx.doi.org/10.13140/RG.2.2.11351.79522 |

This repository implements a hybrid generative model+ combining a Conditional Variational Autoencoder (CVAE), a Conditional Generative Adversarial Network (CGAN), and a Random Forest Classifier (RFC). The hybrid model is designed for complex classification tasks by leveraging latent feature extraction (CVAE), synthetic data generation (CGAN), and robust classification (RFC).

---

Features

1. Conditional Variational Autoencoder (CVAE):

   • Learns a latent representation of input data.
   • Uses a custom loss function combining reconstruction loss and KL divergence.

2. Conditional Generative Adversarial Network (CGAN):

   • Generates high-quality synthetic data conditioned on class labels.

3. Random Forest Classifier:

   • Trained on a combination of real and synthetic features to enhance classification performance.

4. Parallel Model Training:

   • The CVAE and CGAN models are trained concurrently using Python's multiprocessing module.

5. Hyperparameter Tuning:

   • Utilizes Keras Tuner for optimizing the CVAE and CGAN hyperparameters.
   • Random Forest is tuned using GridSearchCV.

---

Installation

Prerequisites

• Python 3.7+
• TensorFlow 2.8+
• Scikit-learn
• Keras Tuner
• Matplotlib
• Seaborn
• Multiprocessing

Install Required Packages

pip install tensorflow keras-tuner scikit-learn matplotlib seaborn

---

---

Install the Package

the package can be installed the package via PyPI:

pip install LatentBoostClassifier

Alternatively, install directly from GitHub:

pip install git+https://github.com/AliBavarchee/LatentBoostClassifier.git

---

Usage

---

Using the Package

After installation, the package can be imported and used like this:

from LatentBoostClassifier import parallel_train, visualize_hybrid_model

# Train the hybrid model
best_cvae, best_cgan_generator, best_rf_model = parallel_train(X_train, Y_train, X_test, Y_test)

# Visualize results
visualize_hybrid_model(best_cvae, best_cgan_generator, best_rf_model, X_test, Y_test, X_train, Y_train)

---

Training the Hybrid Model

Train the CVAE, CGAN, and Random Forest models using your dataset:

# Import the training function
from LatentBoostClassifier import parallel_train

# Define training and testing datasets
X_train, Y_train = ...  # Load or preprocess your training data
X_test, Y_test = ...    # Load or preprocess your test data

# Train the hybrid model
best_cvae, best_cgan_generator, best_rf_model = parallel_train(X_train, Y_train, X_test, Y_test)

---

Visualizing Results

Evaluate and visualize the performance of the hybrid model:

from LatentBoostClassifier import visualize_hybrid_model

# Visualize results
visualize_hybrid_model(best_cvae, best_cgan_generator, best_rf_model, X_test, Y_test, X_train, Y_train)

---

Code Overview

1. CVAE

• Purpose: Extract latent features from input data.
• Key Components:
  • Custom loss function: Combines reconstruction loss and KL divergence.
  • Encoder-Decoder architecture.
• Hyperparameter Tuning:
  • Latent dimensions.
  • Dense layer units.
  • Learning rate.

2. CGAN

• Purpose: Generate synthetic data conditioned on class labels.
• Key Components:
  • Generator: Produces synthetic samples.
  • Discriminator: Distinguishes real and synthetic samples.
• Hyperparameter Tuning:
  • Latent dimensions.
  • Dense layer units.
  • Learning rate.

3. Random Forest

• Purpose: Classify combined real and synthetic data.
• Key Components:
  • Trained on latent features (CVAE) and synthetic data (CGAN).
  • Grid search for hyperparameter optimization.

---

Visualization

Classification Performance

• Confusion Matrix: Visualize actual vs. predicted class distributions.
• Classification Report: Display precision, recall, F1-score, and accuracy metrics.

Latent Space

• Use t-SNE to visualize the CVAE's latent space.

Synthetic Data

• Compare real and synthetic data distributions using KDE plots.

ROC Curve

• Evaluate the Random Forest classifier with an ROC curve and compute AUC.

---

Key Functions

Model Training

parallel_train(X_train, Y_train, X_test, Y_test)

• Trains the CVAE, CGAN, and Random Forest models in parallel.

Visualization

visualize_hybrid_model(best_cvae, best_cgan_generator, best_rf_model, X_test, Y_test, X_train, Y_train)

• Visualizes the hybrid model's results, including latent space and classification performance.

---

Examples

Training Example

# Example Dataset
X_train = np.random.rand(80, 10)
Y_train = np.random.randint(0, 2, size=(80, 1))
X_test = np.random.rand(20, 10)
Y_test = np.random.randint(0, 2, size=(20, 1))

# Train the hybrid model
best_cvae, best_cgan_generator, best_rf_model = parallel_train(X_train, Y_train, X_test, Y_test)

Visualization Example

# Visualize results
visualize_hybrid_model(best_cvae, best_cgan_generator, best_rf_model, X_test, Y_test, X_train, Y_train)

---

Contributions

Feel free to fork, improve, and submit a pull request! For major changes, please open an issue first to discuss what you'd like to change.

---

License

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

=============================================

=============================================

| https://github.com/AliBavarchee/ |

| https://www.linkedin.com/in/ali-bavarchee-qip/ |