quantum-melbourne / qiskit-hackathon-22

A repository for Qiskit Hackathon Melbourne (July 4-7, 2022)

5 stars 2 forks source link

Approximating large-depth QNNs with a Quantum Tangent Kernel #10

Open maiyuren opened 2 years ago

maiyuren commented 2 years ago

Abstract

The analysis of over-parameterised Artificial Neural Networks (ANNs) reveal that the optimisation (training) process only slightly changes the parameters of the model. This allows one to approximate an ANN – up to first order – as a linear model with a defining tangent kernel. In this project we will explore where this regime occurs for variational quantum algorithms and explicitly compute the associated quantum tangent kernel to finally approximate a Quantum Neural Network (QNN) as a kernel method.

Description

We will be using Python and the Qiskit library to code up our quantum machine learning model. More information about the quantum tangent kernel can be found:

Shirai, N., Kubo, K., Mitarai, K. and Fujii, K., 2021. Quantum tangent kernel. arXiv preprint arXiv:2111.02951.

Analysis will be done on simulators and if we have time we can see how our models perform with real devices.

If you don't know anything about machine learning (ML) don't worry! We will go through the ML required on the first day.

Members

Qiskit Coach: @maiyuren

Deliverable

Analysis of parameter evolution in variational algorithms with larger depth circuits
Kernel approximation of QNNs in the large depth regime

GitHub repo

Coming soon!

ParthDatwani commented 2 years ago

Hi I would Like to be a part of ur team

footscrazy commented 2 years ago

I would also like to be a part of your team please

Edit: sorry I can no longer take part - good luck!

UoMzhhhh commented 2 years ago

This is Hang, I would like to be one of this group.

JoshDuff commented 2 years ago

I would love to be a part of this team!

JoshDuff commented 2 years ago

QNN to Kernel Project Presentation.pptx