Deep learning is revolutionizing computing for an ever-increasing range of applications, from natural language processing to particle physics to cancer diagnosis. These advances have been made possible by a combination of algorithmic design and dedicated hardware development. Quantum computing, while more nascent, is experiencing a similar trajectory, with a rapidly closing gap between current hardware and the scale required for practical implementation of quantum algorithms. But we are still extremely far away from a full-scale quantum computer that can implement gate-based computer architectures. Such architectures require quantum error correction to make the system robust against noise, which remains outside the reach of existing quantum technology. This project aims to develop a new approach to quantum computation by adopting concepts from the field of machine learning. In contrast to conventional approaches where computation is decomposed into logical gates, the investigators will focus on quantum computing architectures inspired by machine learning and deep learning to implement quantum protocols that are naturally efficient and robust to noise. These architectures are ideally suited to maximize the computational capabilities of currently available noisy quantum processors because machine learning algorithms can be trained using efficient methods such as back-propagation. The project represents a highly multi-disciplinary effort that combines quantum hardware development with algorithms and computer architecture design to create quantum protocols and devices that can be leveraged for near-term application in quantum simulation, machine learning, optimization, and quantum communication. Success of the project could open a completely new approach to quantum computing that enables currently available quantum hardware to efficiently solve problems in a broad range of fields such as medicine, biology, nuclear physics, and fundamental quantum science. The program also entails a strong outreach effort that integrates education at the high school, undergraduate, and graduate levels with public education through a series of YouTube educational modules.
Integrated quantum photonics enables dynamic, high-fidelity generation and manipulation of quantum states of light, and is therefore a natural platform with which to develop chip-based quantum machine learning architectures. Leveraging both the versatility of neural networks and the computational complexity of quantum optics, the program develops chip-based deep quantum optical neural networks for applications in quantum computation, simulation, communication, machine learning, and beyond. Taking inspiration from the burgeoning field of neural networks, this hardware platform combines semiconductor quantum light sources (input encoding) with dynamically reconfigurable linear optical circuitry (matrix multiplication) and strong single photon nonlinearities (the quantum neuron), to develop a new paradigm for next generation quantum processors. In parallel, the theory effort will develop a robust numerical platform to simulate quantum machine learning protocols based on the hardware platform and design new protocols for multiple applications including image and pattern recognition, optimization, and quantum communication. The strong collaborative interactions between hardware and theory will thus be leveraged to develop an entirely new arsenal of protocols that exploit the unique physical properties of photons.
This project is jointly funded by Quantum Leap Big Idea Program and the Division of Electrical, Communications, and Cyber Systems in the Directorate for Engineering.
This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
