Direct Kernel Optimization: Efficient Design for Opto-Electronic Convolutional Neural Networks

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Almuallem, Ali, Weligampola, Harshana, Gnanasambandam, Abhiram, Xu, Wei, Godaliyadda, Dilshan, Sheikh, Hamid R., Chan, Stanley H., Guo, Qi
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910020754472960
author Almuallem, Ali
Weligampola, Harshana
Gnanasambandam, Abhiram
Xu, Wei
Godaliyadda, Dilshan
Sheikh, Hamid R.
Chan, Stanley H.
Guo, Qi
author_facet Almuallem, Ali
Weligampola, Harshana
Gnanasambandam, Abhiram
Xu, Wei
Godaliyadda, Dilshan
Sheikh, Hamid R.
Chan, Stanley H.
Guo, Qi
contents Hybrid opto-electronic neural networks combine optical front-ends with electronic back-ends to perform vision tasks, but joint end-to-end (E2E) optimization of optical and electronic components is computationally expensive due to large parameter spaces and repeated optical convolutions. We propose Direct Kernel Optimization (DKO), a two-stage training framework that first trains a conventional electronic CNN and then synthesizes optical kernels to replicate the first-layer convolutional filters, reducing optimization dimensionality and avoiding hefty simulated optical convolutions during optimization. We evaluate DKO in simulation on a monocular depth estimation model and show that it achieves twice the accuracy of E2E training under equal computational budgets while reducing training time. Given the substantial computational challenges of optimizing hybrid opto-electronic systems, our results position DKO as a scalable optimization approach to train and realize these systems.
format Preprint
id arxiv_https___arxiv_org_abs_2511_02065
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Direct Kernel Optimization: Efficient Design for Opto-Electronic Convolutional Neural Networks
Almuallem, Ali
Weligampola, Harshana
Gnanasambandam, Abhiram
Xu, Wei
Godaliyadda, Dilshan
Sheikh, Hamid R.
Chan, Stanley H.
Guo, Qi
Image and Video Processing
Computer Vision and Pattern Recognition
Hybrid opto-electronic neural networks combine optical front-ends with electronic back-ends to perform vision tasks, but joint end-to-end (E2E) optimization of optical and electronic components is computationally expensive due to large parameter spaces and repeated optical convolutions. We propose Direct Kernel Optimization (DKO), a two-stage training framework that first trains a conventional electronic CNN and then synthesizes optical kernels to replicate the first-layer convolutional filters, reducing optimization dimensionality and avoiding hefty simulated optical convolutions during optimization. We evaluate DKO in simulation on a monocular depth estimation model and show that it achieves twice the accuracy of E2E training under equal computational budgets while reducing training time. Given the substantial computational challenges of optimizing hybrid opto-electronic systems, our results position DKO as a scalable optimization approach to train and realize these systems.
title Direct Kernel Optimization: Efficient Design for Opto-Electronic Convolutional Neural Networks
topic Image and Video Processing
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.02065