Tools

Quantum Support Vector Machines (QSVM)

1: Data Selection

  1. Navigate to the "Data Selection" stage in the simulation.
  2. Choose a dataset from the available options (e.g., Linear Dataset, Circular Dataset, Xor Dataset).
  3. Observe the "Dataset Distribution" scatter plot. The objective is to find a boundary separating the classes (green vs. red).
  4. Click Next Step.

2: Classical SVM

  1. In the "Classical SVM" stage, review the prompt. This step trains a classical Support Vector Machine using a Linear Kernel.
  2. Click the Run Classical SVM button.
  3. Once completed, view the results:
    • Identify the straight line (hyperplane) separating the classes.
    • Note the Accuracy (e.g., 98%).
    • Highlighted points denote the Support Vectors defining the margin.
  4. Click Next Step.

3: Quantum Feature Map

  1. In the "Quantum Feature Map" stage, configure the map parameters to map data into a high-dimensional quantum state space:
    • Feature Map Type: Set to ZZFeatureMap.
    • Circuit Depth (Repetitions): Set the slider (e.g., 2).
    • Entanglement: Select from Linear, Full, or Circular (e.g., Full).
  2. Observe the generated "Quantum Feature Map Circuit (ZZFeatureMap)" showing Hadamard and Phase Gates.
  3. Review the "Kernel Matrix Heatmap" resulting from mapping to the quantum space.
  4. Click Next Step.

4: Quantum SVM

  1. In the "Quantum SVM" stage, read the background on how the quantum computer calculation matrix is fed to a classical SVM optimizer.

  2. Click the Run QSVM button.

  3. Once completed, view the results:

    • See the "Quantum SVM Decision Boundary", which finds a linear hyperplane in a high-dimensional quantum space, translating to a non-linear complex boundary in the 2D space.
    • Note the Accuracy (e.g., 99%).

  4. Click Next Step.

5: Comparison

  1. Compare the Classical SVM and Quantum SVM decision boundaries side-by-side.
  2. Review the accuracies (e.g., Classical 98% vs Quantum 99%).
  3. Read the Conclusion. For linearly separable data, models perform well across the board, proving that QSVM acts properly on basic problems.
  4. If needed, click Restart to try another dataset or configuration.