Deep quantum graph dreaming: deciphering neural network insights into quantum experiments
Despite their promise to facilitate new scientific discoveries, the opaqueness of neural networks presents a challenge in interpreting the logic behind their findings. Here, we use a eXplainable-AI technique called inception or deep dreaming , which has been invented in machine learning for computer...
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IOP Publishing
2024-01-01
|
Series: | Machine Learning: Science and Technology |
Subjects: | |
Online Access: | https://doi.org/10.1088/2632-2153/ad2628 |
_version_ | 1797306833826742272 |
---|---|
author | Tareq Jaouni Sören Arlt Carlos Ruiz-Gonzalez Ebrahim Karimi Xuemei Gu Mario Krenn |
author_facet | Tareq Jaouni Sören Arlt Carlos Ruiz-Gonzalez Ebrahim Karimi Xuemei Gu Mario Krenn |
author_sort | Tareq Jaouni |
collection | DOAJ |
description | Despite their promise to facilitate new scientific discoveries, the opaqueness of neural networks presents a challenge in interpreting the logic behind their findings. Here, we use a eXplainable-AI technique called inception or deep dreaming , which has been invented in machine learning for computer vision. We use this technique to explore what neural networks learn about quantum optics experiments. Our story begins by training deep neural networks on the properties of quantum systems. Once trained, we ‘invert’ the neural network—effectively asking how it imagines a quantum system with a specific property, and how it would continuously modify the quantum system to change a property. We find that the network can shift the initial distribution of properties of the quantum system, and we can conceptualize the learned strategies of the neural network. Interestingly, we find that, in the first layers, the neural network identifies simple properties, while in the deeper ones, it can identify complex quantum structures and even quantum entanglement. This is in reminiscence of long-understood properties known in computer vision, which we now identify in a complex natural science task. Our approach could be useful in a more interpretable way to develop new advanced AI-based scientific discovery techniques in quantum physics. |
first_indexed | 2024-03-08T00:47:33Z |
format | Article |
id | doaj.art-e7684506beda4ff1b4a8f0e2106ccdc2 |
institution | Directory Open Access Journal |
issn | 2632-2153 |
language | English |
last_indexed | 2024-03-08T00:47:33Z |
publishDate | 2024-01-01 |
publisher | IOP Publishing |
record_format | Article |
series | Machine Learning: Science and Technology |
spelling | doaj.art-e7684506beda4ff1b4a8f0e2106ccdc22024-02-15T07:18:49ZengIOP PublishingMachine Learning: Science and Technology2632-21532024-01-015101502910.1088/2632-2153/ad2628Deep quantum graph dreaming: deciphering neural network insights into quantum experimentsTareq Jaouni0https://orcid.org/0009-0006-5661-2403Sören Arlt1Carlos Ruiz-Gonzalez2https://orcid.org/0000-0003-0545-360XEbrahim Karimi3https://orcid.org/0000-0002-8168-7304Xuemei Gu4Mario Krenn5Max Planck Institute for the Science of Light , Erlangen, Germany; Nexus for Quantum Technologies, University of Ottawa , K1N 6N5, ON, Ottawa, CanadaMax Planck Institute for the Science of Light , Erlangen, GermanyMax Planck Institute for the Science of Light , Erlangen, GermanyMax Planck Institute for the Science of Light , Erlangen, Germany; Nexus for Quantum Technologies, University of Ottawa , K1N 6N5, ON, Ottawa, CanadaMax Planck Institute for the Science of Light , Erlangen, GermanyMax Planck Institute for the Science of Light , Erlangen, GermanyDespite their promise to facilitate new scientific discoveries, the opaqueness of neural networks presents a challenge in interpreting the logic behind their findings. Here, we use a eXplainable-AI technique called inception or deep dreaming , which has been invented in machine learning for computer vision. We use this technique to explore what neural networks learn about quantum optics experiments. Our story begins by training deep neural networks on the properties of quantum systems. Once trained, we ‘invert’ the neural network—effectively asking how it imagines a quantum system with a specific property, and how it would continuously modify the quantum system to change a property. We find that the network can shift the initial distribution of properties of the quantum system, and we can conceptualize the learned strategies of the neural network. Interestingly, we find that, in the first layers, the neural network identifies simple properties, while in the deeper ones, it can identify complex quantum structures and even quantum entanglement. This is in reminiscence of long-understood properties known in computer vision, which we now identify in a complex natural science task. Our approach could be useful in a more interpretable way to develop new advanced AI-based scientific discovery techniques in quantum physics.https://doi.org/10.1088/2632-2153/ad2628neural network interpretabilitydeep dreamingquantum physics |
spellingShingle | Tareq Jaouni Sören Arlt Carlos Ruiz-Gonzalez Ebrahim Karimi Xuemei Gu Mario Krenn Deep quantum graph dreaming: deciphering neural network insights into quantum experiments Machine Learning: Science and Technology neural network interpretability deep dreaming quantum physics |
title | Deep quantum graph dreaming: deciphering neural network insights into quantum experiments |
title_full | Deep quantum graph dreaming: deciphering neural network insights into quantum experiments |
title_fullStr | Deep quantum graph dreaming: deciphering neural network insights into quantum experiments |
title_full_unstemmed | Deep quantum graph dreaming: deciphering neural network insights into quantum experiments |
title_short | Deep quantum graph dreaming: deciphering neural network insights into quantum experiments |
title_sort | deep quantum graph dreaming deciphering neural network insights into quantum experiments |
topic | neural network interpretability deep dreaming quantum physics |
url | https://doi.org/10.1088/2632-2153/ad2628 |
work_keys_str_mv | AT tareqjaouni deepquantumgraphdreamingdecipheringneuralnetworkinsightsintoquantumexperiments AT sorenarlt deepquantumgraphdreamingdecipheringneuralnetworkinsightsintoquantumexperiments AT carlosruizgonzalez deepquantumgraphdreamingdecipheringneuralnetworkinsightsintoquantumexperiments AT ebrahimkarimi deepquantumgraphdreamingdecipheringneuralnetworkinsightsintoquantumexperiments AT xuemeigu deepquantumgraphdreamingdecipheringneuralnetworkinsightsintoquantumexperiments AT mariokrenn deepquantumgraphdreamingdecipheringneuralnetworkinsightsintoquantumexperiments |