Quantum Reservoir Computing

  • 9 November 2022
  • 13:00 - 14:50
  • DAV.0.29

Dr. Alexandre Zagoskin, Department of Physics, º¬Ðß²ÝÊÓƵ

Reservoir computing can be considered as a special case of the neural network approach. Here the network is split in two parts: the reservoir and the readout. The reservoir is a fixed, nonlinear, dissipative physical system with a large space of available states. The readout is a simple neural network trained on the output states of the reservoir. The reservoir plays the role of a nonlinear filter, which efficiently separates the outputs corresponding to different classes of inputs, thus reducing the requirements to the readout network. The approach is especially well suited for temporal/sequential data processing, such as time series prediction and natural speech recognition.

The use of a quantum reservoir has several advantages: it allows quantum inputs, it may be composed of a smaller number of units, and the requirements to the reservoir are much easier to meet than those to, e.g., a quantum annealer.

I will give an overview of the current status of quantum reservoir computing and possible directions of its development for the immediate future.

Contact and booking details

Booking required?
No