⚛️ Classical vs Quantum SVM (PennyLane)

An interactive Streamlit web application comparing the performance of Classical SVM (RBF Kernel) and Quantum SVM (PennyLane-based Quantum Kernel) across various datasets, with adjustable parameters, visualizations, and evaluation metrics.

---

🚀 Features

🧠 Model Comparison

• Classical SVM (RBF Kernel):
  Implements a standard radial basis function kernel: [ K(x_i, x_j) = \exp(-\gamma |x_i - x_j|^2) ]

• Quantum SVM:
  Uses PennyLane’s default.qubit simulator to encode data into quantum states using parameterized rotations: [ K(x, y) = |\langle \psi(x) | \psi(y) \rangle|^2 ]

📊 Evaluation Metrics

• Accuracy
• Precision
• Recall
• F1 Score
• Processing Time

🧩 Interactive Dataset Options

• Upload your own CSV file
• Preloaded sample datasets:
  • Iris
  • Glass
  • Social Network Ads

🧮 Preprocessing & Configurations

• Automatic label encoding and one-hot encoding
• StandardScaler normalization
• Optional PCA (Principal Component Analysis)
• Adjustable test split ratio and quantum training sample limits

📈 Visualization Tools

• Confusion matrices for both models
• Decision boundaries (for 2D data)
• Metric comparison bar charts
• Metric variation across multiple dataset splits
• Cross-dataset summary with trend plots and bar comparisons

---

🧰 Tech Stack

Component	Technology
Frontend/UI	Streamlit
Classical ML	Scikit-learn (SVM, PCA, Metrics)
Quantum ML	PennyLane (default.qubit backend)
Visualization	Matplotlib, Seaborn
Data Handling	Pandas, NumPy

---

📂 Project Structure

📦 classical-vs-quantum-svm
│
├── app.py                           # Main Streamlit application
├── datasets/                        # Dataset storage
│   ├── iris.csv
│   ├── glass.csv
│   ├── social_network_ads.csv
│   └── dataset_summary_results.csv  # Optional summary data
│
├── README.md                        # Project documentation

---

🧮 Adjustable Parameters (Sidebar Controls)

Parameter	Description
PCA Components	Reduce feature dimensions for visualization and quantum kernel stability
Test Size	Fraction of dataset used for testing
Max Quantum Samples	Limits dataset size for quantum circuit simulation
Quantum Limit Factor	Simulates hardware constraints by reducing training data
Dataset Choice	Select from preloaded datasets or upload your own CSV

---

🧩 Outputs & Visualizations

Section	Description
Metric Dashboard	Displays accuracy, precision, recall, F1, and time for both models
Confusion Matrices	Heatmaps for both Classical and Quantum predictions
Decision Boundary	For 2D data, shows model separation visually
Metric Comparison Chart	Side-by-side bar chart comparison
Metric Variation Across Splits	Shows metric fluctuations across 10 sample splits
Cross-Dataset Summary	Aggregates results across datasets for deeper insight
Trend Plots & Bar Charts	Compare Classical vs Quantum across dataset sizes

---

📚 Conceptual Overview

This project demonstrates:

• The practical differences between classical kernel methods and quantum-enhanced kernels.
• How quantum circuits can be used to construct data-dependent kernels for classification tasks.
• The trade-offs between accuracy and computational cost under varying dataset sizes and quantum limitations.

It serves as a foundation for Quantum-Assisted Machine Learning (QAML) exploration and benchmarking.

---

🧠 Research Relevance

This project replicates and extends experiments inspired by the IEEE paper:
“Comparative Analysis of a Quantum SVM With an Optimized Kernel Versus Classical SVMs.”

It provides a real-time, visual, and interactive environment to analyze both models’ behavior under identical conditions.

---

📜 License

This project is released under the MIT License.
You are free to use, modify, and distribute it with attribution.

---

👨‍💻 Author

PlatinumManX
🎓 Engineering Student | 💻 Technical Game Dev Enthusiast | ⚛️ Quantum ML Explorer
📫 Connect: GitHub Profile