Quantum SVMs [Help Needed for implementation 🙏]

[Priyansh]

Hello! I am doing some experiments related to Quantum SVMs of 3 types:
I have implemented the first one QK-SVM on my own (with help of some resources). I would greatly appreciate some help for implementation of the other two types (QV-SVM & QVK-SVM)

I am also attaching circuit images for reference.

1)Quantum Kernel SVM

2)Quantum Variational SVM

3)Quantum Variational Kernel SVM

Here’s the code for the QKSVM

n_qubits= 8

dev = qml.device("lightning.qubit",wires = n_qubits)

@qml.qnode(dev)
def qksvm_kernel_cirq(a,b):
  qml.AngleEmbedding(a,wires = range(n_qubits))
  qml.adjoint(qml.AngleEmbedding(b,wires = range(n_qubits)))
  return qml.probs(wires = range(n_qubits))

def quantum_kernel_pca(A,B):
  return np.array([[qksvm_kernel_cirq(a,b)[0] for b in B] for a in A])

svm = SVC(kernel = quantum_kernel_pca).fit(xs_tr, y_tr)

from sklearn.metrics import accuracy_score
print(accuracy_score(svm.predict(xs_test), y_test))

This is the accuracy

0.9615384615384616

And, finally, make sure to include the versions of your packages. Specifically, show us the output of qml.about().

Name: PennyLane
Version: 0.36.0
Summary: PennyLane is a cross-platform Python library for quantum computing, quantum machine learning, and quantum chemistry. Train a quantum computer the same way as a neural network.
Home-page: https://github.com/PennyLaneAI/pennylane
Author: 
Author-email: 
License: Apache License 2.0
Location: /usr/local/lib/python3.10/dist-packages
Requires: appdirs, autograd, autoray, cachetools, networkx, numpy, pennylane-lightning, requests, rustworkx, scipy, semantic-version, toml, typing-extensions
Required-by: PennyLane_Lightning

Platform info:           Linux-6.1.58+-x86_64-with-glibc2.35
Python version:          3.10.12
Numpy version:           1.25.2
Scipy version:           1.11.4
Installed devices:
- lightning.qubit (PennyLane_Lightning-0.36.0)
- default.clifford (PennyLane-0.36.0)
- default.gaussian (PennyLane-0.36.0)
- default.mixed (PennyLane-0.36.0)
- default.qubit (PennyLane-0.36.0)
- default.qubit.autograd (PennyLane-0.36.0)
- default.qubit.jax (PennyLane-0.36.0)
- default.qubit.legacy (PennyLane-0.36.0)
- default.qubit.tf (PennyLane-0.36.0)
- default.qubit.torch (PennyLane-0.36.0)
- default.qutrit (PennyLane-0.36.0)
- default.qutrit.mixed (PennyLane-0.36.0)
- null.qubit (PennyLane-0.36.0)

Reference paper

[Guillermo_Alonso]

Hi, thanks for the question! It is not an easy change what you comment. Let me try to give you some guidance anyway

First of all, you must be familiar with variational circuits, for which, you can orient yourself from this demo.

Secondary , you should add some variational parameters in qksvm_kernel_cirq that you will train.
The idea of variational SVM is to perform a two-stage training. You will have to define a loss function that is the accuracy of the model you get. That is, within that function, you have to do the SVM training itself.

In your code, accuracy_score can be treated as the new error function that depends of the parameters. It may seem confusing but you will see that it makes sense!

[Priyansh]

Thanks for your guidance, I really appreciate it.

Sure, I will perform some experiments to strengthen my concepts.
I will share my learnings and try to perform the implementation myself.

I’ll get back soon.
Thanks again @Guillermo_Alonso.