A Deep Learning and Signal Processing Architecture Using Frequency-Encoded RF Photonics
Deep neural networks have become ubiquitous due to their ability to perform arbitrary tasks more accurately than manually-crafted systems. This ability has created a substantial demand for more complex models processing larger amounts of data. However, the traditional computing architecture has re...
Main Author: | |
---|---|
Other Authors: | |
Format: | Thesis |
Published: |
Massachusetts Institute of Technology
2023
|
Online Access: | https://hdl.handle.net/1721.1/147409 |
_version_ | 1811095881427976192 |
---|---|
author | Davis, Ronald A. |
author2 | Englund, Dirk |
author_facet | Englund, Dirk Davis, Ronald A. |
author_sort | Davis, Ronald A. |
collection | MIT |
description | Deep neural networks have become ubiquitous due to their ability to perform arbitrary tasks more accurately than manually-crafted systems. This ability has created a substantial demand for more complex models processing larger amounts of data. However, the traditional computing architecture has reached a bottleneck in processing performance due to data movement. Considerable efforts have been made to create custom hardware to accelerate deep neural network training and inference. Among these efforts are optical neural networks, which have been a promising approach that excel at linear operations but struggle with nonlinear implementations. Here, we propose our multiplicative analog frequency transform optical neural network (MAFT-ONN) that computes matrix products using frequency-encoded signals and implements the nonlinearity for each layer using a single Mach-Zhender modulator. We experimentally demonstrate a 3-layer DNN for inference of MNIST digits, showing a scalable, fully analog front-to-end ONN. This architecture is also the first deep neural network hardware accelerator that is suited for direct inference of time-based signals without digitization. |
first_indexed | 2024-09-23T16:32:16Z |
format | Thesis |
id | mit-1721.1/147409 |
institution | Massachusetts Institute of Technology |
last_indexed | 2024-09-23T16:32:16Z |
publishDate | 2023 |
publisher | Massachusetts Institute of Technology |
record_format | dspace |
spelling | mit-1721.1/1474092023-01-20T03:03:15Z A Deep Learning and Signal Processing Architecture Using Frequency-Encoded RF Photonics Davis, Ronald A. Englund, Dirk Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Deep neural networks have become ubiquitous due to their ability to perform arbitrary tasks more accurately than manually-crafted systems. This ability has created a substantial demand for more complex models processing larger amounts of data. However, the traditional computing architecture has reached a bottleneck in processing performance due to data movement. Considerable efforts have been made to create custom hardware to accelerate deep neural network training and inference. Among these efforts are optical neural networks, which have been a promising approach that excel at linear operations but struggle with nonlinear implementations. Here, we propose our multiplicative analog frequency transform optical neural network (MAFT-ONN) that computes matrix products using frequency-encoded signals and implements the nonlinearity for each layer using a single Mach-Zhender modulator. We experimentally demonstrate a 3-layer DNN for inference of MNIST digits, showing a scalable, fully analog front-to-end ONN. This architecture is also the first deep neural network hardware accelerator that is suited for direct inference of time-based signals without digitization. S.M. 2023-01-19T19:48:16Z 2023-01-19T19:48:16Z 2022-09 2022-10-19T18:57:00.354Z Thesis https://hdl.handle.net/1721.1/147409 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
spellingShingle | Davis, Ronald A. A Deep Learning and Signal Processing Architecture Using Frequency-Encoded RF Photonics |
title | A Deep Learning and Signal Processing Architecture Using
Frequency-Encoded RF Photonics |
title_full | A Deep Learning and Signal Processing Architecture Using
Frequency-Encoded RF Photonics |
title_fullStr | A Deep Learning and Signal Processing Architecture Using
Frequency-Encoded RF Photonics |
title_full_unstemmed | A Deep Learning and Signal Processing Architecture Using
Frequency-Encoded RF Photonics |
title_short | A Deep Learning and Signal Processing Architecture Using
Frequency-Encoded RF Photonics |
title_sort | deep learning and signal processing architecture using frequency encoded rf photonics |
url | https://hdl.handle.net/1721.1/147409 |
work_keys_str_mv | AT davisronalda adeeplearningandsignalprocessingarchitectureusingfrequencyencodedrfphotonics AT davisronalda deeplearningandsignalprocessingarchitectureusingfrequencyencodedrfphotonics |