An on-chip photonic deep neural network for image classification

Ashtiani, Farshid; Geers, Alexander J.; Aflatouni, Firooz

doi:10.1038/s41586-022-04714-0

Cited by 218 publications

(123 citation statements)

References 41 publications

Supporting

Mentioning

121

Contrasting

Unclassified

Order By: Relevance

“…In addition, while multiplexing techniques address the mapping of neurons' input nodes and synapses, the nonlinear functions, albeit demonstrated in [55], remains challenging for integrated platforms. They can be potentially achieved via either highly-nonlinear optical materials/structures, or electrooptic devices [73] employed in the interfaces of ONNs. While analog optics features potentially much higher computing power and energy efficiency, they are inherently limited in terms of flexibility and versatility in contrast to digital electronics based on Von Neumann structures with distributed processors and memories.…”

Section: Discussionmentioning

confidence: 99%

Section: (C))mentioning

confidence: 99%

“…The reported ONNs based SDM also include those that adopt an array of grating couplers [73], spatially distributed phase-change material (PCM) meshes [38], verticalcavity surface-emitting lasers (VCSELs) arrays [74] and so on. In [73], an integrated photonic deep neural network (Fig. 3(M)) with optoelectronic nonlinear activation functions was demonstrated, capable of directly processing optical waves impinging on an array of grating couplers and fulfilling image classification.…”

Section: (C))mentioning

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: Discussionmentioning

confidence: 99%

Section: (C))mentioning

confidence: 99%

Section: (C))mentioning

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

“…Some architectures suffer from component scaling challenges scaling faster than the weight matrix size [2]. Others suffer from addressing a specific problem, lacking a high-speed reconfiguration solution for the same architecture, either due to design and materials choice [1], or due to the fixed size of the neural network [26]. The optimal solution should have the flexibility to integrate both low-energy consumption memories, such as electronically-controlled photonic random access memories (P-RAM) using phase change materials (PCM) [1,27,28], as well as high-speed modulators and photodiodes, to fulfill the progress of PTC as the next AI accelerator [29].…”

Section: Introductionmentioning

confidence: 99%

High-density Integrated Photonic Tensor Processing Unit with a Matrix Multiply Compiler

Peserico

Ahmed

et al. 2022

Preprint

View full text Add to dashboard Cite

Machine learning and artificial intelligence (AI) is becoming a ubiquitous technology of the looming industry-4.0 era. However, progress of adopting intelligent automation of systems is limited by hardware overhead such as throughput, power consumption, and latency. At a conceptual level, electronics is at the end of its scaling law and alternative accelerators are sought after. Optical co-processors offer a high-degree of algorithmic homomorphism to implement general matrix-matrix multiplication operations via on-the-fly multiplication performed by electro-optic components, and accumulation operations performed by photodetectors. However, recent emerging photonic AI engines are cumbersome to program, follow overhead-heavy scaling laws, or rely on discretely-packaged photonic components, all reducing performance. Here we introduce a photonic tensor core processor featuring a complete chip-integrated multiply-accumulate engine on the photonic circuits, signal parallelism via wavelength division multiplexing, and tunable 5-bit electronic programmable weights, all integrated into a small-formfactor silicon photonics platform. This approach shows a high throughput-efficiency AI engine (0.21 TOPS/W). Using this stand-alone hybrid photonic-electronic machine learning accelerator, we demonstrate versatile applications including low-error image filtering and edge detection, augmented reality, and high-accuracy machine learning image classification tasks. This hybrid electronic-photonic tensor processor offers a versatile low-overhead and compact formfactor solution for extreme edge AI applications, but also can be utilized for in-cloud machine learning training tasks.

show abstract

“…(L) An integrated diffractive neural network[68]. (M) An integrated photonic deep neural network based on spatially distributed array of input grating couplers[73]. (N) A spatially distributed 16-node on-chip RC[77].…”

mentioning

confidence: 99%

Photonic multiplexing methods for optical neuromorphic computing

Moss¹

2022

Preprint

View full text Add to dashboard Cite

<p> 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. </p>

show abstract

An on-chip photonic deep neural network for image classification

Cited by 218 publications

References 41 publications

Photonic Multiplexing Techniques for Optical Neuromorphic Computing

Photonic Multiplexing Techniques for Optical Neuromorphic Computing

High-density Integrated Photonic Tensor Processing Unit with a Matrix Multiply Compiler

Photonic multiplexing methods for optical neuromorphic computing

Contact Info

Product

Resources

About