top of page

Primer on silicon neuromorphic photonic processors


Challenges in Microelectronic Computers: It acknowledges the limitations of current microelectronic computers in meeting today's information processing needs. The paper posits that unconventional computers employing alternative processing models, like neural network models, are essential for addressing these challenges​​.


Evolution of Computing: The paper notes that despite significant advances in digital microelectronics since the 1960s, there are computational problems that remain out of reach. This has led to a resurgence in unconventional computing approaches, including neuromorphic electronics and radio frequency (RF) photonics​​.


Neuromorphic Photonics: It highlights the potential of neuromorphic photonics to integrate the physical models of optoelectronic systems with abstract models of neural networks. This integration is seen as a promising avenue for machine information processing, capable of operating on sub-nanosecond timescales and applicable to a range of tasks from mathematical programming to real-time control​​.


Role of Silicon Photonics: The paper emphasizes the critical role of silicon photonics in making large-scale and low-cost photonic systems a reality. It underscores that the integration of neural network models and silicon photonics can have significant impacts on machine information processing​​.


Silicon Neuromorphic Photonic Processor: The paper discusses the vision for a silicon neuromorphic photonic processor. It is envisioned as a complement to digital microelectronic computing or as an alternate platform for new applications like nonlinear programming or wideband radio signal processing​​.


Deep Learning and Neural Networks: The growing importance of neural network models, particularly deep learning, in modern machine learning is highlighted. The paper notes that these models have become indispensable in various applications due to algorithmic innovations, the proliferation of data on the internet, and new hardware like GPUs​​.


Transition to Optical Computing: The paper reflects on the transition from initial attempts at optical computing to contemporary photonic information processing. It recognizes the multifaceted nature of optical computing and the learning curve involved in identifying viable approaches and avoiding pitfalls​​.


Barriers and Potential of Silicon Photonics: It discusses the barriers faced by silicon photonics in the past, such as a lack of a low-cost platform, and how recent developments have overcome these barriers, positioning silicon photonics as a key player in the future of photonic information processing​​.


Advancements in Photonic Computing Models: The paper also explores various models of computing with light, including quantum silicon photonics, photonic reservoir computing, and RF photonics, each leveraging different properties of lightwaves​​​​.


This provides a thorough examination of the state and potential of silicon neuromorphic photonic processors, offering insights into their architecture, applications, and the broader context of their development within the fields of photonics and computing.



bottom of page