Prospects and applications of photonic neural networks

Huang, Chaoran; Sorger, Volker J.; Miscuglio, Mario; Al-Qadasi, Mohammed; Mukherjee, Avilash; Lampe, Lutz; Nichols, Mitchell; Tait, Alexander N.; Lima, Thomas Ferreira de; Márquez, Bicky A.; Wang, Jiahui; Chrostowski, Lukas; Fok, Mable P.; Brunner, Daniel; Fan, Shanhui; Shekhar, Sudip; Prucnal, Paul R.; Shastri, Bhavin J.

doi:10.1080/23746149.2021.1981155

Cited by 81 publications

(56 citation statements)

References 177 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Spatial-division multiplexing is a fundamental approach to boost the computing parallelism and enhance the overall computing speed, as has been widely used in traditional digital computing systems [14][15][16][17][18][19][20][21][22][23][24][25]. ONNs based on SDM feature architectures where the input nodes X and/or weighted synapses W are mapped onto the spatial division.…”

Section: Onns Based On Space-division Multiplexingmentioning

confidence: 99%

“…The unique advantages of light, such as its ultrawide bandwidths of up to 10's of THz, the low propagation loss and the inherent nature of its analog architecture, make optical neuromorphic computing hardware promising to address the challenges faced by their electronic counterparts [14][15][16][17][18][19][20][21][22][23][24][25]. Ultimately hybrid opto-electronic computing hardware that leverage the broad bandwidths of optics without sacrificing the flexibility of digital electronics may provide the ideal solution.…”

Section: Introductionmentioning

confidence: 99%

“…Recently, the advances in ONNs have been reviewed from a number of different perspectives [14][15][16][17][18][19][20][21][22][23][24][25], including introducing: optical field interferences for visual computing applications [18], integrated neuromorphic systems and the underlying hardware for implementing weighted interconnects and neurons [19], the training methods [21], energy consumption [22] and prospects and applications of ONNs [23]. Here, we review the most recent advances of ONNs from the perspective of the fundamental photonic multiplexing techniques that offer physical parallelism for the implementation of ONNs.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Photonic Multiplexing Techniques for Optical Neuromorphic Computing

Moss¹

2022

Preprint

View full text Add to dashboard Cite

The simultaneous advances in artificial neural networks and photonic integration technologies have spurred extensive research in optical computing and optical neural networks (ONNs). The potential to simultaneously exploit multiple physical dimensions of time, wavelength and space give ONNs the ability to achieve computing operations with high parallelism and large-data throughput. Different photonic multiplexing techniques based on these multiple degrees of freedom have enabled ONNs with large-scale interconnectivity and linear computing functions. Here, we review the recent advances of ONNs based on different approaches to photonic multiplexing, and present our outlook on key technologies needed to further advance these photonic multiplexing/hybrid-multiplexing techniques of ONNs.

show abstract

Section: Onns Based On Space-division Multiplexingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Photonic Multiplexing Techniques for Optical Neuromorphic Computing

Moss¹

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…ANN based on optical devices, the so-called optical neural network (ONN), can overcome the aforementioned defects. Compared with the electronics counterpart, optical transmission inherently has the advantages of high speed, large bandwidth, low delay, high parallelism, low energy consumption, and immunity to electro-magnetic interference (EMI) [5]. Therefore, ONN can take advantage of photonic technology to break the bottleneck of conventional electrical ANN.…”

Section: Introductionmentioning

confidence: 99%

Silicon-Based Metastructure Optical Scattering Multiply–Accumulate Computation Chip

et al. 2022

View full text Add to dashboard Cite

Optical neural networks (ONN) have become the most promising solution to replacing electronic neural networks, which have the advantages of large bandwidth, low energy consumption, strong parallel processing ability, and super high speed. Silicon-based micro-nano integrated photonic platforms have demonstrated good compatibility with complementary metal oxide semiconductor (CMOS) processing. Therefore, without completely changing the existing silicon-based fabrication technology, optoelectronic hybrid devices or all-optical devices of better performance can be achieved on such platforms. To meet the requirements of smaller size and higher integration for silicon photonic computing, the topology of a four-channel coarse wavelength division multiplexer (CWDM) and an optical scattering unit (OSU) are inversely designed and optimized by Lumerical software. Due to the random optical power splitting ratio and incoherency, the intensities of different input signals from CWDM can be weighted and summed directly by the subsequent OSU to accomplish arbitrary multiply–accumulate (MAC) operations, therefore supplying the core foundation for scattering ONN architecture.

show abstract

“…T HE field of artificial intelligence has undergone major changes in the last decade, which can be attributed to the success of deep learning when working with large datasets and advances in computational hardware [1]. However, as the post-Moore era draws near, conventional computers are not projected to keep up with the increasing demands [2]. Neuromorphic engineering promises to achieve new hardware architectures that match the distributed nature of machine learning (ML) algorithms, leading the way towards faster and more sustainable computation [1], [2].…”

Section: Introductionmentioning

confidence: 99%

Data-driven Modeling of Mach-Zehnder Interferometer-based Optical Matrix Multipliers

Ali¹,

Yan²,

Ding³

et al. 2022

Preprint

View full text Add to dashboard Cite

Photonic integrated circuits are facilitating the development of optical neural networks, which have the potential to be both faster and more energy efficient than their electronic counterparts since optical signals are especially well-suited for implementing matrix multiplications. However, accurate programming of photonic chips for optical matrix multiplication remains a difficult challenge. Here, we describe both simple analytical models and data-driven models for offline training of optical matrix multipliers. We train and evaluate the models using experimental data obtained from a fabricated chip featuring a Mach-Zehnder interferometer mesh implementing 3-by-3 matrix multiplication. The neural network-based models outperform the simple physics-based models in terms of prediction error. Furthermore, the neural network models are also able to predict the spectral variations in the matrix weights for up to 100 frequency channels covering the C-band. The use of neural network models for programming the chip for optical matrix multiplication yields increased performance on multiple machine learning tasks.

show abstract

Prospects and applications of photonic neural networks

Cited by 81 publications

References 177 publications

Photonic Multiplexing Techniques for Optical Neuromorphic Computing

Photonic Multiplexing Techniques for Optical Neuromorphic Computing

Silicon-Based Metastructure Optical Scattering Multiply–Accumulate Computation Chip

Data-driven Modeling of Mach-Zehnder Interferometer-based Optical Matrix Multipliers

Contact Info

Product

Resources

About