Hakan Türeci is a Professor of Electrical and Computer Engineering at Princeton University. His research primarily focuses on non-equilibrium dynamics in classical and quantum optical collective systems, with a recent focus on quantum electrodynamics of superconducting devices, the physics of computation, and computing with physical systems. 

Recent advances in machine learning show that practically any controllable physical system responding to physical stimuli can perform computation. This opens new opportunities to use Physical Neural Networks (PNNs) and helps deepen our understanding of biological neural circuits. To fully exploit this potential, PNNs must be trained with awareness of the physical nature of signal and noise – where signal is defined relative to the task.

In his talk at the Max Planck Institute for the Science of Light (MPL), Hakan Türeci addressed the question of how physical systems can be used as neural networks. A key point here is the role of noise. “A physical system is characterized by the existence of noise. Hence, in practice, a physical system’s behavior may significantly deviate from theoretical description,” says Türeci. It is therefore essential to characterize the relationship between signal-to-noise behavior and the performance of systems in various machine learning tasks. In particular, he demonstrated that a low signal-to-noise ratio and quantum correlations can increase the learning ability of physical neural networks. He backed up these theoretical predictions in his presentation with impressive experimental results on a superconducting quantum processor. His work also has important implications for applications in the fields of quantum machine learning and sensing.

You can watch Prof. Türeci’s talk here:
https://www.youtube.com/watch?v=CJtIMx44HH4&list=PL6yOOrXfatatYcdjK_N90smVlWbVLrfj6

Past talks are available on our YouTube channel in the DLS playlist.

If you don't want to miss out on any future DLS events, please sign up for our DLS newsletter here.