SUANPAN: Scalable Photonic Linear Vector Machine

Ziyue Yang,Chen Li,Yuqia Ran,Yongzhuo Li,Xue Feng,Kaiyu Cui,Fang Liu,Hao Sun,Wei Zhang,Yu Ye,Fei Qiao,Cun-Zheng Ning,Jiaxing Wang,Connie J.Chang-Hasnain,Yidong Huang

2024-10-31

Abstract:Photonic linear operation is a promising approach to handle the extensive vector multiplications in artificial intelligence techniques due to the natural bosonic parallelism and high-speed information transmission of photonics. Although it is believed that maximizing the interaction of the light beams is necessary to fully utilize the parallelism and tremendous efforts have been made in past decades, the achieved dimensionality of vector-matrix multiplication is very limited due to the difficulty of scaling up a tightly interconnected or highly coupled optical system. Additionally, there is still a lack of a universal photonic computing architecture that can be readily merged with existing computing system to meet the computing power demand of AI techniques. Here, we propose a programmable and reconfigurable photonic linear vector machine to perform only the inner product of two vectors, formed by a series of independent basic computing units, while each unit is just one pair of light-emitter and photodetector. Since there is no interaction among light beams inside, extreme scalability could be achieved by simply duplicating the independent basic computing unit while there is no requirement of large-scale analog-to-digital converter and digital-to-analog converter arrays. Our architecture is inspired by the traditional Chinese Suanpan or abacus and thus is denoted as photonic SUANPAN. As a proof of principle, SUANPAN architecture is implemented with an 8*8 vertical cavity surface emission laser array and an 8*8 MoTe2 two-dimensional material photodetector array. We believe that our proposed photonic SUANPAN is capable of serving as a fundamental linear vector machine that can be readily merged with existing electronic digital computing system and is potential to enhance the computing power for future various AI applications.

Optics

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on two aspects:

1. **Scalability problem of high - dimensional vector operations**: In traditional photonic computing architectures, when performing high - dimensional vector - matrix multiplications, due to the difficulty in large - scale expansion of the interactions between light beams, the achievable dimensions are very limited. Moreover, the existing photonic computing architectures lack a general - purpose architecture that can be seamlessly integrated with the existing electronic digital computing systems to meet the computing power requirements of artificial intelligence technologies.

2. **Data conversion problem in optoelectronic hybrid computing**: In optoelectronic hybrid computing, the conversion between analog data and digital data is the main limiting factor. In particular, the use of large - scale analog - to - digital converter (ADC) and digital - to - analog converter (DAC) arrays not only increases the complexity of the system but also limits the scalability and speed of computing.

To solve these problems, the author proposes a programmable and reconfigurable photonic linear vector machine - **SUANPAN**, whose design is inspired by the traditional Chinese abacus. SUANPAN performs the inner product operation of two vectors through a series of independent basic computing units (BEADs). Each basic unit consists of a pair of emitters and photodetectors. This design avoids the interactions between light beams, thereby achieving extremely high scalability, and only one ADC is required to convert the total analog current into a digital signal, greatly simplifying the system structure.

Specifically, the main features of SUANPAN include:
- **Scalability**: By simply replicating independent basic computing units (BEADs), the computing scale can be easily expanded without introducing additional losses or errors.
- **Low complexity**: Since each BEAD only needs to control the on - off state of the emitter and the position of the photodetector, complex DAC and ADC arrays are not required.
- **Flexibility**: It can flexibly adapt to the requirements of different computing tasks by reconfiguring the number and arrangement of BEADs.

Through these designs, SUANPAN can significantly improve computing efficiency and scalability while maintaining high computing accuracy, providing strong support for various future AI applications.