Optical neural network

Last updated
Optical logic gate with diffractive neural network.png
Microwave diffractive ANN.png
Schematic of an optical neural network that functions as a logic gate (above) and its implementation in microwave frequencies (below). The intermediate diffractive metasurfaces function as hidden layers. [1]

An optical neural network is a physical implementation of an artificial neural network with optical components. Early optical neural networks used a photorefractive Volume hologram to interconnect arrays of input neurons to arrays of output with synaptic weights in proportion to the multiplexed hologram's strength. [2] Volume holograms were further multiplexed using spectral hole burning to add one dimension of wavelength to space to achieve four dimensional interconnects of two dimensional arrays of neural inputs and outputs. [3] This research led to extensive research on alternative methods using the strength of the optical interconnect for implementing neuronal communications. [4]

Contents

Some artificial neural networks that have been implemented as optical neural networks include the Hopfield neural network [5] and the Kohonen self-organizing map with liquid crystal spatial light modulators [6] Optical neural networks can also be based on the principles of neuromorphic engineering, creating neuromorphic photonic systems. Typically, these systems encode information in the networks using spikes, mimicking the functionality of spiking neural networks in optical and photonic hardware. Photonic devices that have demonstrated neuromorphic functionalities include (among others) vertical-cavity surface-emitting lasers, [7] [8] integrated photonic modulators, [9] optoelectronic systems based on superconducting Josephson junctions [10] or systems based on resonant tunnelling diodes. [11]

Electrochemical vs. optical neural networks

Biological neural networks function on an electrochemical basis, while optical neural networks use electromagnetic waves. Optical interfaces to biological neural networks can be created with optogenetics, but is not the same as an optical neural networks. In biological neural networks there exist a lot of different mechanisms for dynamically changing the state of the neurons, these include short-term and long-term synaptic plasticity. Synaptic plasticity is among the electrophysiological phenomena used to control the efficiency of synaptic transmission, long-term for learning and memory, and short-term for short transient changes in synaptic transmission efficiency. Implementing this with optical components is difficult, and ideally requires advanced photonic materials. Properties that might be desirable in photonic materials for optical neural networks include the ability to change their efficiency of transmitting light, based on the intensity of incoming light.

Rising Era of Optical Neural Networks

With the increasing significance of computer vision in various domains, the computational cost of these tasks has increased, making it more important to develop the new approaches of the processing acceleration. Optical computing has emerged as a potential alternative to GPU acceleration for modern neural networks, particularly considering the looming obsolescence of Moore's Law. Consequently, optical neural networks have garnered increased attention in the research community. Presently, two primary methods of optical neural computing are under research: silicon photonics-based and free-space optics. Each approach has its benefits and drawbacks; while silicon photonics may offer superior speed, it lacks the massive parallelism that free-space optics can deliver. Given the substantial parallelism capabilities of free-space optics, researchers have focused on taking advantage of it. One implementation, proposed by Lin et al., [12] involves the training and fabrication of phase masks for a handwritten digit classifier. By stacking 3D-printed phase masks, light passing through the fabricated network can be read by a photodetector array of ten detectors, each representing a digit class ranging from 1 to 10. Although this network can achieve terahertz-range classification, it lacks flexibility, as the phase masks are fabricated for a specific task and cannot be retrained. An alternative method for classification in free-space optics, introduced by Cahng et al., [13] employs a 4F system that is based on the convolution theorem to perform convolution operations. This system uses two lenses to execute the Fourier transforms of the convolution operation, enabling passive conversion into the Fourier domain without power consumption or latency. However, the convolution operation kernels in this implementation are also fabricated phase masks, limiting the device's functionality to specific convolutional layers of the network only. In contrast, Li et al. [14] proposed a technique involving kernel tiling to use the parallelism of the 4F system while using a Digital Micromirror Device (DMD) instead of a phase mask. This approach allows users to upload various kernels into the 4F system and execute the entire network's inference on a single device. Unfortunately, modern neural networks are not designed for the 4F systems, as they were primarily developed during the CPU/GPU era. Mostly because they tend to use a lower resolution and a high number of channels in their feature maps.

Other Implementations

In 2007 there was one model of Optical Neural Network: the Programmable Optical Array/Analogic Computer (POAC). It had been implemented in the year 2000 and reported based on modified Joint Fourier Transform Correlator (JTC) and Bacteriorhodopsin (BR) as a holographic optical memory. Full parallelism, large array size and the speed of light are three promises offered by POAC to implement an optical CNN. They had been investigated during the last years with their practical limitations and considerations yielding the design of the first portable POAC version.

The practical details – hardware (optical setups) and software (optical templates) – were published. However, POAC is a general purpose and programmable array computer that has a wide range of applications including:

Progress in the 2020s

Taichi from Tsinghua University in Beijing is a hybrid ONN that combines the power efficiency and parallelism of optical diffraction and the configurability of optical interference. Taichi offers 13.96 million parameters. Taichi avoids the high error rates that afflict deep (multi-layer) networks by combining clusters of fewer-layer diffractive units with arrays of interferometers for reconfigurable computation. Its encoding protocol divides large network models into sub-models that can be distributed across multiple chiplets in parallel. [15]

Taichi achieved 91.89% accuracy in tests with the Omniglot database. It was also used to generate music Bach and generate images the styles of Van Gogh and Munch. [15]

The developers claimed energy efficiency of up to 160 trillion operations second-1 watt-1 and an area efficiency of 880 trillion multiply-accumulate operations mm-2 or 103 more energy efficient than the NVIDIA H100, and 102 times more energy efficient and 10 times more area efficient than previous ONNs. [15]

Time dimension has recently been introduced into diffrative nueral network by fs laser lithography of perovskite hydration. The temporal behaviour of the neuron can be modulated by the fs laser at the nanoscale, enabling a programmable holographic neural network with temporal evolution functionality, i.e., the functionality can change with time under the hydration stimuli. An in-memory temporal inference functionality was demonstrated to mimic the function evolution of the human brain,i.e.,the functionality can change from simple digit image classification to more complicated digit and clothing product image classification with time. This is the first time of introducting time dimension into the optical neural netwrok, laying a foundation for future brain-like photonic chip development. [16]

See also

Related Research Articles

<span class="mw-page-title-main">Photonic crystal</span> Periodic optical nanostructure that affects the motion of photons

A photonic crystal is an optical nanostructure in which the refractive index changes periodically. This affects the propagation of light in the same way that the structure of natural crystals gives rise to X-ray diffraction and that the atomic lattices of semiconductors affect their conductivity of electrons. Photonic crystals occur in nature in the form of structural coloration and animal reflectors, and, as artificially produced, promise to be useful in a range of applications.

Neuromorphic computing is an approach to computing that is inspired by the structure and function of the human brain. A neuromorphic computer/chip is any device that uses physical artificial neurons to do computations. In recent times, the term neuromorphic has been used to describe analog, digital, mixed-mode analog/digital VLSI, and software systems that implement models of neural systems. Recent advances have even discovered ways to mimic the human nervous system through liquid solutions of chemical systems.

Holonomic brain theory is a branch of neuroscience investigating the idea that human consciousness is formed by quantum effects in or between brain cells. Holonomic refers to representations in a Hilbert phase space defined by both spectral and space-time coordinates. Holonomic brain theory is opposed by traditional neuroscience, which investigates the brain's behavior by looking at patterns of neurons and the surrounding chemistry.

Optical computing or photonic computing uses light waves produced by lasers or incoherent sources for data processing, data storage or data communication for computing. For decades, photons have shown promise to enable a higher bandwidth than the electrons used in conventional computers.

<span class="mw-page-title-main">Optical vortex</span> Optical phenomenon

An optical vortex is a zero of an optical field; a point of zero intensity. The term is also used to describe a beam of light that has such a zero in it. The study of these phenomena is known as singular optics.

<span class="mw-page-title-main">Extraordinary optical transmission</span>

Extraordinary optical transmission (EOT) is the phenomenon of greatly enhanced transmission of light through a subwavelength aperture in an otherwise opaque metallic film which has been patterned with a regularly repeating periodic structure. Generally when light of a certain wavelength falls on a subwavelength aperture, it is diffracted isotropically in all directions evenly, with minimal far-field transmission. This is the understanding from classical aperture theory as described by Bethe. In EOT however, the regularly repeating structure enables much higher transmission efficiency to occur, up to several orders of magnitude greater than that predicted by classical aperture theory. It was first described in 1998.

Digital holography is the acquisition and processing of holograms with a digital sensor array, typically a CCD camera or a similar device. Image rendering, or reconstruction of object data is performed numerically from digitized interferograms. Digital holography offers a means of measuring optical phase data and typically delivers three-dimensional surface or optical thickness images. Several recording and processing schemes have been developed to assess optical wave characteristics such as amplitude, phase, and polarization state, which make digital holography a very powerful method for metrology applications .

Phased-array optics is the technology of controlling the phase and amplitude of light waves transmitting, reflecting, or captured (received) by a two-dimensional surface using adjustable surface elements. An optical phased array (OPA) is the optical analog of a radio-wave phased array. By dynamically controlling the optical properties of a surface on a microscopic scale, it is possible to steer the direction of light beams, or the view direction of sensors, without any moving parts. Phased-array beam steering is used for optical switching and multiplexing in optoelectronic devices and for aiming laser beams on a macroscopic scale.

Computer-generated holography (CGH) is a technique that uses computer algorithms to generate holograms. It involves generating holographic interference patterns. A computer-generated hologram can be displayed on a dynamic holographic display, or it can be printed onto a mask or film using lithography. When a hologram is printed onto a mask or film, it is then illuminated by a coherent light source to display the holographic images.

<span class="mw-page-title-main">Spiking neural network</span> Artificial neural network that mimics neurons

Spiking neural networks (SNNs) are artificial neural networks (ANN) that more closely mimic natural neural networks. These models leverage timing of discrete spikes as the main information carrier.

<span class="mw-page-title-main">Microlens</span> Small lens, generally with a diameter less than a millimetre

A microlens is a small lens, generally with a diameter less than a millimetre (mm) and often as small as 10 micrometres (μm). The small sizes of the lenses means that a simple design can give good optical quality but sometimes unwanted effects arise due to optical diffraction at the small features. A typical microlens may be a single element with one plane surface and one spherical convex surface to refract the light. Because micro-lenses are so small, the substrate that supports them is usually thicker than the lens and this has to be taken into account in the design. More sophisticated lenses may use aspherical surfaces and others may use several layers of optical material to achieve their design performance.

<span class="mw-page-title-main">Acousto-optic programmable dispersive filter</span>

An acousto-optic programmable dispersive filter (AOPDF) is a special type of collinear-beam acousto-optic modulator capable of shaping spectral phase and amplitude of ultrashort laser pulses. AOPDF was invented by Pierre Tournois. Typically, quartz crystals are used for the fabrication of the AOPDFs operating in the UV spectral domain, paratellurite crystals are used in the visible and the NIR and calomel in the MIR (3–20 μm). Recently introduced lithium niobate crystals allow for high-repetition rate operation (> 100 kHz) owing to their high acoustic velocity. The AOPDF is also used for the active control of the carrier-envelope phase of few-cycle optical pulses, as a part of pulse-measurement schemes and multi-dimensional spectroscopy techniques. Although sharing a lot in principle of operation with an acousto-optic tunable filter, the AOPDF should not be confused with it, since in the former the tunable parameter is the transfer function and in the latter it is the impulse response.

Microstructured optical fibers (MOF) are optical fiber waveguides where guiding is obtained through manipulation of waveguide structure rather than its index of refraction.

Photonic molecules are a form of matter in which photons bind together to form "molecules". They were first predicted in 2007. Photonic molecules are formed when individual (massless) photons "interact with each other so strongly that they act as though they have mass". In an alternative definition, photons confined to two or more coupled optical cavities also reproduce the physics of interacting atomic energy levels, and have been termed as photonic molecules.

<span class="mw-page-title-main">Electromagnetic metasurface</span>

An electromagnetic metasurface refers to a kind of artificial sheet material with sub-wavelength features. Metasurfaces can be either structured or unstructured with subwavelength-scaled patterns.

An AI accelerator, deep learning processor or neural processing unit (NPU) is a class of specialized hardware accelerator or computer system designed to accelerate artificial intelligence and machine learning applications, including artificial neural networks and computer vision. Typical applications include algorithms for robotics, Internet of Things, and other data-intensive or sensor-driven tasks. They are often manycore designs and generally focus on low-precision arithmetic, novel dataflow architectures or in-memory computing capability. As of 2024, a typical AI integrated circuit chip contains tens of billions of MOSFETs.

<span class="mw-page-title-main">Virtually imaged phased array</span> Dispersive optical device

A virtually imaged phased array (VIPA) is an angular dispersive device that, like a prism or a diffraction grating, splits light into its spectral components. The device works almost independently of polarization. In contrast to prisms or regular diffraction gratings, the VIPA has a much higher angular dispersion but has a smaller free spectral range. This aspect is similar to that of an Echelle grating, since it also uses high diffraction orders. To overcome this disadvantage, the VIPA can be combined with a diffraction grating. The VIPA is a compact spectral disperser with high wavelength resolving power.

Cornelia Denz is a German Professor of Physics at the University of Münster. She works in nonlinear optics and nanophotonics, and is a Fellow of The Optical Society and The European Optical Society. Denz is the current president of the PTB.

Deep learning in photoacoustic imaging

Deep learning in photoacoustic imaging combines the hybrid imaging modality of photoacoustic imaging (PA) with the rapidly evolving field of deep learning. Photoacoustic imaging is based on the photoacoustic effect, in which optical absorption causes a rise in temperature, which causes a subsequent rise in pressure via thermo-elastic expansion. This pressure rise propagates through the tissue and is sensed via ultrasonic transducers. Due to the proportionality between the optical absorption, the rise in temperature, and the rise in pressure, the ultrasound pressure wave signal can be used to quantify the original optical energy deposition within the tissue.

In photonics, a meta-waveguide is a physical structures that guides electromagnetic waves with engineered functional subwavelength structures. Meta-waveguides are the result of combining the fields of metamaterials and metasurfaces into integrated optics. The design of the subwavelength architecture allows exotic waveguiding phenomena to be explored.

References

  1. Qian, Chao; Lin, Xiao; Lin, Xiaobin; Xu, Jian; Sun, Yang; Li, Erping; Zhang, Baile; Chen, Hongsheng (2020). "Performing optical logic operations by a diffractive neural network". Light: Science & Applications . 9 (59): 59. Bibcode:2020LSA.....9...59Q. doi:10.1038/s41377-020-0303-2. PMC   7154031 . PMID   32337023.
  2. Wagner K, Psaltis D (1988). "Adaptive optical networks using photorefractive crystals". Appl. Opt. 27 (9): 1752–1759. Bibcode:1988ApOpt..27.1752P. doi:10.1364/AO.27.001752. PMID   20531647.
  3. Weverka R, Wagner K, Saffman M (1991). "Fully interconnected, two-dimensional neural arrays using wavelength-multiplexed volume holograms". Optics Letters. 16 (11): 826–828. Bibcode:1991OptL...16..826W. doi:10.1364/OL.16.000826. PMID   19776798.
  4. Wagner K, Psaltis D (1993). "Optical neural networks: an introduction by the feature editors". Appl. Opt. 32 (8): 1261–1263. Bibcode:1993ApOpt..32.1261W. doi:10.1364/AO.32.001261. PMID   20820259.
  5. Ramachandran R, Gunasekaran N (2000). "Optical Implementation of Two Dimensional Bipolar Hopfield Model Neural Network (Scientific Note)" (PDF). Proceedings-National Science Council Republic of China Part a Physical Science and Engineering. 24 (1): 73–8. Archived from the original (PDF) on 12 October 2004.
  6. Duvillier J, Killinger M, Heggarty K, Yao K, de Bougrenet de la Tocnaye JL (January 1994). "All-optical implementation of a self-organizing map: a preliminary approach". Applied Optics. 33 (2): 258–66. Bibcode:1994ApOpt..33..258D. doi:10.1364/AO.33.000258. PMID   20862015.
  7. Hejda M, Robertson J, Bueno J, Alanis J, Hurtado A (2021-06-01). "Neuromorphic encoding of image pixel data into rate-coded optical spike trains with a photonic VCSEL-neuron". APL Photonics. 6 (6): 060802. Bibcode:2021APLP....6f0802H. doi: 10.1063/5.0048674 . ISSN   2378-0967.
  8. Robertson J, Hejda M, Bueno J, Hurtado A (April 2020). "Ultrafast optical integration and pattern classification for neuromorphic photonics based on spiking VCSEL neurons". Scientific Reports. 10 (1): 6098. Bibcode:2020NatSR..10.6098R. doi:10.1038/s41598-020-62945-5. PMC   7142074 . PMID   32269249.
  9. George JK, Mehrabian A, Amin R, Meng J, de Lima TF, Tait AN, et al. (February 2019). "Neuromorphic photonics with electro-absorption modulators". Optics Express. 27 (4): 5181–5191. arXiv: 1809.03545 . Bibcode:2019OExpr..27.5181G. doi:10.1364/OE.27.005181. PMID   30876120. S2CID   80625696.
  10. Shainline JM (January 2020). "Fluxonic Processing of Photonic Synapse Events". IEEE Journal of Selected Topics in Quantum Electronics. 26 (1): 1–15. arXiv: 1904.02807 . Bibcode:2020IJSTQ..2627473S. doi:10.1109/JSTQE.2019.2927473. ISSN   1077-260X. S2CID   102352120.
  11. Romeira B, Javaloyes J, Ironside CN, Figueiredo JM, Balle S, Piro O (September 2013). "Excitability and optical pulse generation in semiconductor lasers driven by resonant tunneling diode photo-detectors". Optics Express. 21 (18): 20931–40. Bibcode:2013OExpr..2120931R. doi:10.1364/OE.21.020931. hdl: 10400.1/11954 . PMID   24103966. S2CID   480070.
  12. Lin, Xing; Rivenson, Yair; Yardimci, Nezih T.; Veli, Muhammed; Luo, Yi; Jarrahi, Mona; Ozcan, Aydogan (7 September 2018). "All-optical machine learning using diffractive deep neural networks". Science. 361 (6406): 1004–1008. arXiv: 1804.08711 . Bibcode:2018Sci...361.1004L. doi:10.1126/science.aat8084. PMID   30049787. S2CID   13753997.
  13. Chang, Julie; Sitzmann, Vincent; Dun, Xiong; Heidrich, Wolfgang; Wetzstein, Gordon (17 August 2018). "Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification". Scientific Reports. 8 (1): 12324. Bibcode:2018NatSR...812324C. doi:10.1038/s41598-018-30619-y. PMC   6098044 . PMID   30120316. S2CID   256961403.
  14. Li, Shurui; Miscuglio, Mario; Sorger, Volker J.; Gupta, Puneet (2020). "Channel Tiling for Improved Performance and Accuracy of Optical Neural Network Accelerators". arXiv: 2011.07391 [cs.ET].
  15. 1 2 3 CHOI, CHARLES Q. (12 April 2024). "AI Chip Trims Energy Budget Back by 99+ Percent - IEEE Spectrum". IEEE . Retrieved 2024-04-17.
  16. Zhang, Yinan; Zhu, Shengting; Hu, Jinming; Gu, Min (6 October 2024). "Femtosecond laser direct nanolithography of perovskite hydration for temporally programmable holograms". Nature Communications. 15: 1661. doi:10.1038/s41467-024-51148-5.