Optical reservoir computing for machine learning at the speed of light

用于光速机器学习的光储层计算

• 批准号: 521395-2018
• 负责人: Morandotti, Roberto
• 金额: $ 13.77万
• 依托单位: Institut national de la recherche scientifique
• 依托单位国家: 加拿大
• 项目类别: Strategic Projects - Group
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=694099
• 关键词: Optical; reservoir; computing; machine; learning

In analogy to how today's standard laptops have far exceeded the first punch-card computers, this three-year project targets the advancement of quantum computers capable of succeeding under real-world conditions, based on the reliable control of complex quantum states via 'high-dimensional gate operations'. The proposed research will allow the university team from INRS-EMT and its industrial partners (Passat Ltd. and few-cycle Inc.) to lead international efforts in improving the performance of quantum information processing (QIP), while using practical and low-cost platforms that are highly demanded by industry and academia. These needs can be addressed by quantum optics, as one can store large amounts of information in single photons (particles of light), which can also be transmitted over long distances in standard, low-cost telecommunications fibers. Since, however, photon detection rates are significantly higher in the visible, we will link these two regimes by up-converting photon frequencies. Full control over photons is crucial for QIP, which nevertheless requires quantum gates that are challenging to build, unsuitable for practical use, and still limited in dimensionality (d=4). We aim to boost that to d=8, which is a significant milestone, similar to the information increase from 2D X-Ray images to full 3D CT-scans. We will achieve this goal through three steps: (I) efficient up-conversion of photons through a programmable and compact thermal control unit; (II) universal scheme for arbitrarily mixing the spectral components of classical light with full control of amplitude and phase in the frequency domain; (III) extension of this universal scheme to coherently manipulate photons and arbitrarily mix their frequencies for performing the targeted gate operations on one and two photons. The proposed technology and the training of highly-qualified personnel will contribute to keep Canada at the forefront of the rapidly moving "quantum revolution", thereby enhancing its leading position across critical sectors that vitally require powerful quantum computation platforms, such as advanced environmental sensing, high-precision chemical analysis, and increased security protocols.

与今天的标准笔记本电脑远远超过第一台穿孔卡片计算机类似，这个为期三年的项目旨在推进能够在现实世界条件下成功的量子计算机，其基础是通过“高维门操作”可靠地控制复杂的量子态。拟议中的研究将允许INRS-EMT的大学团队及其工业合作伙伴（帕萨特有限公司和少周期公司）领导国际努力，提高量子信息处理（QIP）的性能，同时使用工业界和学术界高度要求的实用和低成本平台。这些需求可以通过量子光学来解决，因为人们可以将大量信息存储在单个光子（光粒子）中，这些光子也可以在标准的低成本电信光纤中长距离传输。然而，由于光子探测率在可见光中明显更高，我们将通过上转换光子频率将这两个机制联系起来。对光子的完全控制对于QIP至关重要，然而这需要量子门，其构建具有挑战性，不适合实际使用，并且维度仍然有限（d=4）。我们的目标是将其提高到d=8，这是一个重要的里程碑，类似于从2D X射线图像到全3D CT扫描的信息增加。我们将通过三个步骤来实现这一目标：（I）通过可编程和紧凑的热控制单元实现光子的有效上转换;（II）用于任意混合经典光的光谱分量的通用方案，其在频域中完全控制振幅和相位;（三）该通用方案的扩展，以相干地操纵光子并任意混合它们的频率，用于执行目标门操作一个或两个光子。拟议的技术和高素质人才的培训将有助于使加拿大保持在快速发展的“量子革命”的最前沿，从而加强其在关键领域的领先地位，这些领域迫切需要强大的量子计算平台，如先进的环境传感，高精度化学分析和增加的安全协议。