February 26, 2025 11:00 AM
Computing with Physical Systems: Opportunities and Fundamental Limits
Hakan Türeci - Princeton
Recent strides in machine learning have shown that computation can be performed by practically any controllable physical system that responds to physical stimuli encoding data [1]. This perspective opens new frontiers for computational approaches using Physical Neural Networks (PNNs) [2, 3] and provides a framework to deepen our understanding of their biological counterparts—neural circuits in living organisms. To fully leverage this potential, PNNs must be trained with a nuanced awareness of the physical nature of signal and noise, where signal is defined relative to the specific computational task. This perspective aligns closely with approaches to determining fundamental limits in sensing but extends these ideas to a new level to encompass broader computational opportunities. I will share some perspectives on how to approach this new domain of inquiry and some recent results. Within this paradigm, quantum computing emerges as the ultimate quantum noise-dominated limit of physical computation, characterizable by a metric we have recently introduced, the Resolvable Expressive Capacity (REC) [2]. REC quantifies the space of computable functions by a given physical system, and its calculation requires only measured outputs and is, hence, easily characterizable, as was recently demonstrated with optical and quantum systems [2]. I will discuss a recent generalization of the above ideas to inference on streaming data, that enables processing of temporal data over durations unconstrained by the finite coherence times of constituent qubits, without the need for error correction or mitigation [4]. Practical applicability is demonstrated through experiments conducted on IBM's superconducting quantum processors. Finally, I will conclude with results and perspectives on applying this framework to the sensing of high-dimensional weak electromagnetic signals, particularly those originating from measured quantum systems [3]. Based on work with Fangjun Hu, Saeed A. Khan, Gerasimos Angelatos, Marti Vives, Esin Türeci, Graham E. Rowlands, Guilhem J. Ribeill, Nicholas Bronn.
[1] Aspen Center for Physics Winter Conference, Computing with Physical Systems, https://computingwithphysicalsystems.com/2024/
[2] F. Hu et al. ``Tackling Sampling Noise in Physical Systems for Machine Learning Applications: Fundamental Limits and Eigentasks." Phys. Rev. X 13, 041020 (2023).
[3] S. A. Khan et al., ``A neural processing approach to quantum state discrimination", arxiv:2409.03748.
[4] F. Hu et al. ``Overcoming the Coherence Time Barrier in Quantum Machine Learning on Temporal Data", Nature Commun. 15, 7491 (2024).