PennyLaneAI/qml

[DEMO] Quantum multilabel classification with JAX

Closed this issue · 9 comments

poporubeus commented

General information

Name
Francesco Aldo Venturelli (poporubeus).

Affiliation
University of Florence.

Demo information

Title
Quantum multilabel classification with JAX

Abstract
Quantum convolutional neural network (QCNN) for multilabel image classification written by combining Pennylane with JAX.
In this simple tutorial we select four different classes of digits images from the sklearn.datasets.load_digits and we construct a quantum convolutional neural network to train and make the classification of multiclass images.
After model training, we select the last updated and optimal parameters (which correspond to the maximum value of the validation accuracy) and use them to test the model on the unseen test images. To have an idea about all the steps needed to complete the experiment we recommend to have a look at the scripts in the /src folder.
Hope this could be useful for you, feel free to use these codes and make further improvements. With this code we would like to emphasize the ability of making multilabel classification and stop of being constrained by binary classifications!

Relevant links
https://github.com/poporubeus/quantum_machine_learning/tree/main/multilabel_classification/tutorial

poporubeus commented

I submit this simple tutorial on multilabel classification using a QCNN combining Pennylane and JAX together.

ikurecic commented

Thanks a lot for the submission, @poporubeus !
Could I ask you to add a couple more short comments in the Notebook to explain what you're doing and concluding, in the second half? That would be fantastic.

poporubeus commented
poporubeus commented

I have added more comments on the .ipynb. Let me know if they are sufficient. Thanks in advance.

CatalinaAlbornoz commented

Thanks @poporubeus! We'll do another round of review and keep you updated. This may take a few days.

ikurecic commented

Thanks a lot for making the updates, @poporubeus , this is great to see — thanks so much for sharing your code!

You'll be able to already see your demo linked on the PennyLane website today or tomorrow, and we will also post it on social media later on.
Congrats! ☺

poporubeus commented
  • ❤️1
ikurecic commented

Hi @poporubeus , thank you so much for your kind words. :)

I just wanted to quickly follow up to ask you if you'd like our Marketing Team to tag your LinkedIn account (or Twitter account) in a social media post. If that would be fine, please go ahead and share the link with me.

Thank you!

poporubeus commented