Optical neural networks for ultra-fast, low-latency machine intelligence

用于超快、低延迟机器智能的光神经网络

• 批准号: 10043476
• 金额: $ 120.58万
• 依托单位: OXONN LTD
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=10043476
• 关键词: Optical; neural; networks; ultra; fast

Machine intelligence, powered by artificial neural networks, is developing rapidly and entering all aspects of our lives. A novel paradigm in this field is offered by optics, where photons, rather than electrons, play the role of information carriers and computational agents. Optical neural networks promise to enhance both the power efficiency and speed of neural networks by a factor of 1,000-100,000\. Switching from electronic to optical neural networks will make machine intelligence much more accessible and environmentally friendly, thereby delivering direct benefits to society. More productive and energy efficient AI systems will democratise machine learning, making it more of value to socially-vulnerable groups.OxONN in collaboration with the University of Oxford are creating the foundations for the next-generation of advanced optical neural networks. This consortium possesses world-leading expertise in optical computing and has developed game-changing technologies in optical-computing methodologies and hardware. In this project, the consortium will focus on developing the optical implementation of a primary component of any neural network - the optical matrix-vector multiplier, which will be superior to competing products in terms of productivity and scalability, and become OxONN's first MVP. In addition, a conceptually novel deep optics neural network system for computer vision will be developed. This system will allow a neural network to "see" and interpret objects directly, bypassing the need for converting an image into an electronic form. Such a system will have ultra-low latency and find applications in autonomous vehicles, remote sensing and intelligent robotics.This highly innovative project aims to put the UK at the forefront of all-optical computing technology and position our country as the next global technology leader in AI systems for a plethora of applications.

以人工神经网络为动力的机器智能正在迅速发展并进入我们生活的各个方面。光学为该领域提供了一种新颖的范例，其中光子而不是电子发挥着信息载体和计算代理的作用。光学神经网络有望将神经网络的功率效率和速度提高 1,000-100,000 倍。从电子神经网络转向光学神经网络将使机器智能变得更加容易获取和环保，从而为社会带来直接利益。生产力更高、能源效率更高的人工智能系统将使机器学习民主化，使其对社会弱势群体更具价值。OxONN 与牛津大学合作，正在为下一代先进光学神经网络奠定基础。该联盟拥有世界领先的光计算专业知识，并在光计算方法和硬件方面开发了改变游戏规则的技术。在这个项目中，该联盟将专注于开发任何神经网络主要组件的光学实现——光学矩阵矢量乘法器，它将在生产力和可扩展性方面优于竞争产品，并成为 OxONN 的第一个 MVP。此外，还将开发一种概念新颖的计算机视觉深度光学神经网络系统。该系统将允许神经网络直接“看到”并解释物体，无需将图像转换为电子形式。这样的系统将具有超低延迟，并在自动驾驶汽车、遥感和智能机器人领域得到应用。这个高度创新的项目旨在使英国处于全光计算技术的最前沿，并将我国定位为下一个全球技术领导者在人工智能系统中具有大量的应用程序。