Quantum Machine Learning Qiskit

Original repository by Matthew Maccelari

Overview

In this repository, quantum-enhanced support vector machines (QSVMs) are explored by embedding classical data into a high-dimensional Hilbert space via a parameterised quantum feature map and comparing performance against a classical RBF-kernel SVM. The programs:

1. Clean and preprocess a binary classification dataset.
2. Reduce dimensionality with PCA.
3. Train and evaluate a QSVM on both a statevector simulator and real IBM quantum hardware.
4. Benchmark accuracy and runtime against a classical SVM baseline.
5. Visualise quantum kernel matrices as heatmaps.

---

Repository Structure

Quantum-Machine-Learning-REUPLOAD/
├── README.md
├── .env                      ← pinned package versions
├── src/
│   ├── simulatorQML.ipynb    ← QSVM on Qiskit statevector simulator
│   └── hardwareQML.ipynb     ← QSVM on IBM quantum hardware
├── Heatmaps/
│   └── kernel_*.png          ← Saved quantum-kernel heatmaps
└── Dataset/
    └── bots_vs_users.csv     ← Raw binary-classification dataset

---

Requirements

• Python 3.8+
• pandas
• numpy
• scikit-learn
• matplotlib
• Qiskit stack (exact versions)

  pip install \
    qiskit==1.4.3 \
    qiskit-aer==0.11.0 \
    qiskit-ibm-runtime==0.13.1 \
    qiskit-machine-learning==0.8.2

---

Installation & Setup

1. Clone the repo

   git clone https://github.com/your-repo/ELEN4022_LAB3_2025_MATTHEW_MACCELARI.git
   cd ELEN4022_LAB3_2025_MATTHEW_MACCELARI

2. Install dependencies Create and activate the Conda environment from environment.yml:

conda env create -f environment.yml
conda activate elen4022_lab3

3. **Configure IBM Quantum credentials**

   ```bash
   export IBMQ_TOKEN="YOUR_IBM_QUANTUM_API_TOKEN"

or run QiskitRuntimeService.save_account(…) in a Python REPL.

---

Usage

1. Simulator Notebook

Open and run:

jupyter notebook src/simulatorQML.ipynb

This notebook:

• Loads & cleans Dataset/bots_vs_users.csv
• Applies variance filtering and PCA
• Trains QSVM with ZZFeatureMap + FidelityStatevectorKernel on a statevector simulator
• Trains a classical RBF-kernel SVM baseline
• Records accuracies and saves kernel-matrix heatmaps to Heatmaps/

2. Hardware Notebook

Open and run:

jupyter notebook src/hardwareQML.ipynb

This notebook follows the same preprocessing pipeline, then:

1. Builds a 1-repetition ZZFeatureMap circuit on $n$ qubits.
2. Uses QiskitRuntimeService & SamplerV2 to dispatch fidelity-circuit jobs to the least-busy IBM quantum backend.
3. Computes Gram matrices from measurement probabilities of $\lvert0\cdots0\rangle$.
4. Trains a classical SVM with kernel='precomputed' on the real-device kernel.
5. Prints test accuracy and optionally visualises the train-kernel heatmap.

Note: Real-device runs can take several minutes and incur queue times.

---

Results

• Heatmaps: Heatmaps/kernel_n{n}_size{m}.png showing quantum kernel matrices for each PCA dimension $n$ and sample size $m$.
• Accuracy tables: Displayed in both notebooks, comparing QSVM vs classical SVM across experimental conditions.