search

qiskit-community / qiskit-camp-asia-19

31 stars 7 forks source link

Optimization for Reservoir Computing (A method of Neural Networks) #38

Closed trichdop closed 4 years ago

trichdop commented 4 years ago

Abstract

Reservoir computing(RC) is a method of neural networks. The parameters to be optimized are between 0 and 1. Since the probability of the quantum states is always between 0 and 1. Therefore, the quantum computing might be very suitable to optimize the reservoir computing.

Description

In reservoir computing, there are three matrices between input and output data, Win, W and Wout. The values of the elements of Win, W are chosen randomly. Wout is obtained by the matrix inversion. The aim of this Qiskit project is to use the quantum states to optimize the parameters in Win, W and Wout. The detail is illustrated in the following slide. chen

Members

Deliverable

The aim of this pitch in the camp is to build up the algorithm of the reservoir computing using quantum computing for the programming in qiskit.

GitHub repo

trichdop commented 4 years ago

Two papers of Reservoir Computing: [1]Tian-You Cheng, Da-Ya Chou, Ching-Chuan Liu, Ya-Ju Chang, Chii-Chang Chen, "Optical Neural Networks Based on Optical Fiber-communication System," Neurocomputing vol. 364, p. 239, 2019. NEUCOMS0925231219310264-main.pdf [2] Meng-Hua Yen, Ding-Well Liu, Yi-Chia Hsin, Chu-En Lin, Chii-Chang Chen, "Application of the deep learning for the prediction of rainfall in Southern Taiwan," Scientific Reports, vol. 9, p.12774, 2019. srep41598-019-49242-6.pdf