\u3010Speaker\u3011Quoc Hoan TRAN<\/span>@The University of Tokyo<\/span> <\/p>\n \u3010Date\u301110:30-12:00 JST, Wednesday, Jan. 26<\/p>\n \u3010Title\u3011\"Quantum Reservoir Computing - from Classical to Quantum Time Series Processing\"
\u3010Abstract\u3011
Reservoir computing<\/em> is a computational paradigm inspired by the efficient information processing in the brain. It consists of an input-driven high-dimensional dynamical system called a reservoir<\/em>, which can perform complex processing tasks with a learning-free mechanism for the inner parameters of physical systems. For processing time-series data in a temporal task, a reservoir requires memory to keep track of relevant information about the input history. In studying good candidates for the physical implementation of reservoirs, exploiting the quantum nature of physical systems leads to the research field called quantum reservoir computing<\/em> (QRC). QRC has been extensively studied for classical temporal tasks, such as emulating complex spatiotemporal chaos and large-scale nonlinear dynamical systems. Here, the classical input is sequentially encoded into the quantum system while the output is extracted via the measurement after a time evolution of the system. Surprisingly, QRC can also be extended for dealing with quantum data in novel temporal quantum tasks, such as performing quantum tomography of quantum devices that output a series of quantum states depending on the sequence of past input states. This talk will introduce the unified framework and established works in QRC for both classical and quantum time series processing tasks. In light of this, from the information processing perspective, the underlying quantum dynamics behind the learning task\u2019s performance will be analyzed in detail.This talk will be based on the following publications and some unpublished works.
1 - arXiv:2006.08999 (2020)
2 - Phys. Rev. Lett. 127<\/strong>, 260401 (2021)<\/p>\n