Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free Learning
Abstract Plastic self‐adaptation, nonlinear recurrent dynamics and multi‐scale memory are desired features in hardware implementations of neural networks, because they enable them to learn, adapt, and process information similarly to the way biological brains do. In this work, these properties occur...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-01-01
|
| Series: | Advanced Science |
| Subjects: | |
| Online Access: | https://doi.org/10.1002/advs.202404920 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841543173641338880 |
|---|---|
| author | Alessio Lugnan Samarth Aggarwal Frank Brückerhoff‐Plückelmann C. David Wright Wolfram H. P. Pernice Harish Bhaskaran Peter Bienstman |
| author_facet | Alessio Lugnan Samarth Aggarwal Frank Brückerhoff‐Plückelmann C. David Wright Wolfram H. P. Pernice Harish Bhaskaran Peter Bienstman |
| author_sort | Alessio Lugnan |
| collection | DOAJ |
| description | Abstract Plastic self‐adaptation, nonlinear recurrent dynamics and multi‐scale memory are desired features in hardware implementations of neural networks, because they enable them to learn, adapt, and process information similarly to the way biological brains do. In this work, these properties occurring in arrays of photonic neurons are experimentally demonstrated. Importantly, this is realized autonomously in an emergent fashion, without the need for an external controller setting weights and without explicit feedback of a global reward signal. Using a hierarchy of such arrays coupled to a backpropagation‐free training algorithm based on simple logistic regression, a performance of 98.2% is achieved on the MNIST task, a popular benchmark task looking at classification of written digits. The plastic nodes consist of silicon photonics microring resonators covered by a patch of phase‐change material that implements nonvolatile memory. The system is compact, robust, and straightforward to scale up through the use of multiple wavelengths. Moreover, it constitutes a unique platform to test and efficiently implement biologically plausible learning schemes at a high processing speed. |
| format | Article |
| id | doaj-art-dd7d860830b2405c8897f71c7af8a9b7 |
| institution | Kabale University |
| issn | 2198-3844 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Wiley |
| record_format | Article |
| series | Advanced Science |
| spelling | doaj-art-dd7d860830b2405c8897f71c7af8a9b72025-01-13T15:29:43ZengWileyAdvanced Science2198-38442025-01-01122n/an/a10.1002/advs.202404920Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free LearningAlessio Lugnan0Samarth Aggarwal1Frank Brückerhoff‐Plückelmann2C. David Wright3Wolfram H. P. Pernice4Harish Bhaskaran5Peter Bienstman6Photonics Research Group Ghent University‐imec Ghent 9052 BelgiumDepartment of Materials University of Oxford Parks Road Oxford OX1 3PH UKDepartment of Physics CeNTech, University of Münster Heisenbergstraße 48149 Münster GermanyDepartment of Engineering University of Exeter Exeter EX4 4QF UKDepartment of Physics CeNTech, University of Münster Heisenbergstraße 48149 Münster GermanyDepartment of Materials University of Oxford Parks Road Oxford OX1 3PH UKPhotonics Research Group Ghent University‐imec Ghent 9052 BelgiumAbstract Plastic self‐adaptation, nonlinear recurrent dynamics and multi‐scale memory are desired features in hardware implementations of neural networks, because they enable them to learn, adapt, and process information similarly to the way biological brains do. In this work, these properties occurring in arrays of photonic neurons are experimentally demonstrated. Importantly, this is realized autonomously in an emergent fashion, without the need for an external controller setting weights and without explicit feedback of a global reward signal. Using a hierarchy of such arrays coupled to a backpropagation‐free training algorithm based on simple logistic regression, a performance of 98.2% is achieved on the MNIST task, a popular benchmark task looking at classification of written digits. The plastic nodes consist of silicon photonics microring resonators covered by a patch of phase‐change material that implements nonvolatile memory. The system is compact, robust, and straightforward to scale up through the use of multiple wavelengths. Moreover, it constitutes a unique platform to test and efficiently implement biologically plausible learning schemes at a high processing speed.https://doi.org/10.1002/advs.202404920neuromorphic computingmachine learningphase change materialsreservoir computingself‐adapting systemssilicon photonics |
| spellingShingle | Alessio Lugnan Samarth Aggarwal Frank Brückerhoff‐Plückelmann C. David Wright Wolfram H. P. Pernice Harish Bhaskaran Peter Bienstman Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free Learning Advanced Science neuromorphic computing machine learning phase change materials reservoir computing self‐adapting systems silicon photonics |
| title | Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free Learning |
| title_full | Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free Learning |
| title_fullStr | Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free Learning |
| title_full_unstemmed | Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free Learning |
| title_short | Emergent Self‐Adaptation in an Integrated Photonic Neural Network for Backpropagation‐Free Learning |
| title_sort | emergent self adaptation in an integrated photonic neural network for backpropagation free learning |
| topic | neuromorphic computing machine learning phase change materials reservoir computing self‐adapting systems silicon photonics |
| url | https://doi.org/10.1002/advs.202404920 |
| work_keys_str_mv | AT alessiolugnan emergentselfadaptationinanintegratedphotonicneuralnetworkforbackpropagationfreelearning AT samarthaggarwal emergentselfadaptationinanintegratedphotonicneuralnetworkforbackpropagationfreelearning AT frankbruckerhoffpluckelmann emergentselfadaptationinanintegratedphotonicneuralnetworkforbackpropagationfreelearning AT cdavidwright emergentselfadaptationinanintegratedphotonicneuralnetworkforbackpropagationfreelearning AT wolframhppernice emergentselfadaptationinanintegratedphotonicneuralnetworkforbackpropagationfreelearning AT harishbhaskaran emergentselfadaptationinanintegratedphotonicneuralnetworkforbackpropagationfreelearning AT peterbienstman emergentselfadaptationinanintegratedphotonicneuralnetworkforbackpropagationfreelearning |