PhD Scholarships on Photonic Ising Machines:
“Integrated optics implementations” (UGent) and
“Experimental implementations” (VUB)
Closing date: 15th of June 2022
Starting date of scholarships: 1st of September 2022
Three Belgium teams at Vrije Universiteit Brussel (VUB), Université libre de Bruxelles (ULB), Gent University (UGENT) have united their expertise to develop a new type of optical computing devices that can solve hard combinatorial problems.
Such combinatorial problems are commonplace in our society, for instance in logistics, finance or pharmaceutical research. However, for many real-world applications, finding a solution requires high-performance computer clusters that consume large amounts of energy and run for a long time. Our project aims to create a radically new platform of analogue hardware accelerators, so-called Ising machines, that efficiently speed up these computationally difficult tasks in a way unlike any current digital computer. These Ising machines are a newly emerging computational concept and have shown great promise. They are based on the idea that the lowest energy state of a well-constructed dynamical system will encode the solution to the combinatorial problem. However, their implementation is highly challenging due to limited bandwidth, scalability and stability issues. A breakthrough is needed to make them practical for real-world applications. Photonics presents an ideal way to achieve this breakthrough due to its inherent parallelism and high speed. We aim to create accelerators for a broad set of problems, that are orders of magnitude faster and more energy efficient than digital computer and previous experimental implementations.
Our three teams will work on different aspects of this challenge:
ULB will focus on the underlying theory, leading to a deeper understanding of Ising machines and improving their performance. The general approach will be extended to other applications such as mixed integer programming, sampling from the thermal distribution over solutions be studied, and solving real world problems. From the fundamental point of view, we will study the relation between classical and quantum Ising machines, and with other computational models such as adiabatic quantum computing.
VUB will focus on tabletop Ising machines with up to 100.000 spins and orders of magnitude speed up with respect to digital implementations.
UGENT will focus on building integrated versions of photonic Ising machines that contain fewer nodes than the tabletop implementations, but have the potential for much higher speeds and, when produced using volume fabrication, would be cheaper as well.
The researchers will work in close collaboration between the different groups. The selected candidates will carry out their PhD in “cotutelle”, i.e. they will be jointly supervised by two of the groups, will have to carry out part of their research in (at least) two institutions, and will receive a diploma from two institutions. This will ensure efficient flow of ideas, and put the researchers in contact with all aspects of the project. Researchers will be encouraged to work on several aspects of the present project (for instance a specific experimental implementation and the underlying theory).