Random-mapped intensity optical neural network: all-optical two-layer computing for multimodal optical-field inference
Gi-Hyun Go, Doeon Lee, Gookho Song, Mooseok Jang
Abstract
Free-space optical neural networks offer distinct advantages for computational imaging and machine vision because they can compute directly on incident optical fields. However, conventional ONNs composed of cascaded linear optical components are bound to a general linear input-output relation with square-law detection between the input field x and output score y, y=|Tx|2, where at best every complex-valued element of the transmission matrix T is trainable. This restricts each output score to a quadratic form xAx with rank-one decision matrix A=tt. Here, we present a random-mapped intensity optical neural network (RMI-ONN) as an all-optical two-layer computational network that lifts this rank-one limit. We numerically demonstrate that a high-dimensional feature projection by a disordered medium, a programmable nonnegative intensity mask, and segmented spatial power summation together can surpass the rank-one ceiling through the expressivity of higher-rank quadratic decision boundaries. Furthermore, exploiting the vectorial coherent wave-mixing nature of the disordered medium, we experimentally validate multimodal classification of amplitude, phase, and polarization on MNIST, Fashion-MNIST, and Quick Draw with the RMI-ONN, under a single optical configuration across all encoding domains. These results provide a practical and conceptual basis for scalable direct-field optical processors capable of exploiting amplitude, phase, and polarization information within a unified intensity-based inference framework.
Create a lesson
Related papers
Optical Reservoir Computing with Structural Nonlinearity for Forecasting Chaotic Time Series
Leandro R. Venâncio, Jonathan Dong, Mickael Mounaix et al.
Spatially-resolved multiphoton photoemission from a lateral transition metal dichalcogenide heterostructure
Lina Hansen, Paul Martin, Sai Shradha et al.
Exceptional Topological Signatures of Non-Hermitian Photonic Hopf-Link Braids
Samit Kumar Gupta
Comparative study of noise-like pulse emission from Mamyshev oscillator and conventional passively mode-locked fiber laser
Oumaima Ougrige, Florent Bessin, François Sanchez
Static and breathing dynamics of optical bullets in nonlinear GRIN fibers
Sonali Maity, Tiyas Das, Anuj P. Lara et al.
Dispersion-engineered meta-coverslip for multimodal synthetic imaging
Yuchen Ma, Danlin Xu, Xinwei Wang et al.