NFG-Library

A modular, object-oriented Java soft computing library implementing:

• Genetic Algorithms
• Fuzzy Logic Systems
• Neural Networks

The library is supported by real-world case studies, each demonstrating how a soft computing technique can be applied in practice using a clean, reusable API.

---

🧠 Overview

NFG-Library is designed as a generic and extensible soft computing framework.
Each algorithmic paradigm is implemented independently of any specific problem, following clean software architecture principles.

For each project phase, a dedicated case study demonstrates:

• How the library is configured
• How data is preprocessed
• How training and evaluation are performed
• How results are obtained and interpreted

---

🚀 Features

🧬 Genetic Algorithms

• Roulette & rank-based selection
• Crossover and mutation operators
• Fitness function abstraction
• Configurable population size, mutation rate, and generations
• Problem-independent design

---

🌫️ Fuzzy Logic

• Linguistic variables and fuzzy sets
• Membership functions (Triangular, Trapezoidal, Gaussian)
• Mamdani and Sugeno inference engines
• AND / OR operators (T-Norms & S-Norms)
• Defuzzification methods (Centroid, Mean of Maximum)
• Rule-based system with enable/disable and weighting support

---

🧠 Neural Networks

• Feed-forward fully connected neural networks
• Backpropagation training
• Activation functions:
  • ReLU
  • Sigmoid
  • Tanh
  • Linear
• Loss functions:
  • Mean Squared Error (MSE)
  • Binary Cross-Entropy
• Optimizer:
  • Stochastic Gradient Descent (SGD)
• Weight initialization:
  • Xavier initialization
• Configurable architecture and hyperparameters
• Graceful handling of invalid or missing inputs

---

🧪 Case Studies

🔹 Phase 1 — Genetic Algorithms

Problem: Job Scheduling
Goal: Optimize task assignment using evolutionary search
Technique: Genetic Algorithm

---

🔹 Phase 2 — Fuzzy Logic

Problem: Automatic Window Blind Control

Inputs:

• Light Intensity
• Room Temperature

Output:

• Blind Opening Percentage

Technique: Mamdani & Sugeno Fuzzy Inference Systems

---

🔹 Phase 3 — Neural Networks

Problem: Banknote Authentication
Goal: Classify banknotes as genuine or counterfeit

Details:

• Binary classification
• Feed-forward neural network
• ReLU hidden layer + Sigmoid output
• Cross-Entropy loss
• Config-driven architecture

---

📂Project Structure

NFG-Library/
├── src/main/java/          # Main library source code
│   ├── genetic/                       # Genetic Algorithms module with Task Scheduling case study
│   ├── fuzzy/                         # Fuzzy Logic module with case Blind Control case study
│   └── neural_network/                # Neural Networks module with Banknote Authentication case study
├── src/main/resources/                # Configuration files for case studies
├── docs/                              # Documentation
├── pom.xml                            # Maven project file
├── run.sh                             # Script to build and run the project
└── README.md                          # Project overview and instructions

---

▶️ How to Run the Neural Network Case Study (Runnable Demo)

Prerequisites

• Java 17+
• Apache Maven

If Maven is not installed:

sudo apt install maven
# From the project root directory
chmod +x run.sh
./run.sh

The run.sh script will:

1. Build the project using Maven
2. Run the Banknote Authentication neural network case study

---

📚 Technologies Used

• Java 17+
• Maven
• Object-Oriented Design
• Modular Architecture
• Soft Computing Algorithms
• Configuration-driven experiments