Sorry for the basic question.
I have a question about the following qcnn tutorial.
This tutorial includes the following:
「we are not going to train the quantum convolution layer, 」
My understanding is that quantum machine learning generally uses parametrized circuits, and the amount of rotation of the rotation gate is adjusted through learning.
In this tutorial, why is quantum processing merely a pre-processing step and not tuning the parameters of the quantum circuit?
This tutorial demonstrate a “hybrid quantum-classical model” which only use quantum circuit as partly in a machine learning model, where quantum can either be used for the training or merely used for pre/post processing step.
A simulated quantum model requires relatively more computing power to train compared to the classical model, and does not necessarily performs better when replaced a classical counter part, which is a reason sometime quantum circuit are used in a model but does not hold training parameters.
2 Likes
Thank you for answer.
Although learning quantum circuits requires computing power, will it contribute to improving Motel’s performance?
Unfortunately adding a quantum layer is not guaranteed to improve a model’s performance @kou.