Figure 1 | Scientific Reports

Figure 1

From: Optimizing a quantum reservoir computer for time series prediction

Figure 1

(a) In reservoir computing the input data is fed from the input nodes \({\bar{u}}\) with weights \(W^{\mathbf{in }}\) to the network hidden nodes \({\bar{x}}\). The internal immutable dynamics are governed by \(W^{\mathbf{in }}\). Outputs are obtained by weighting the readout nodes \({\bar{z}}\), corresponding to \(\langle Z_i \rangle =\text {Tr}[Z_i \rho ]\) in our setup, with the readout weights \(W^{\mathbf{out }}\). (b) In quantum reservoir computing an input \(s_i\) is fed to a spin by setting its state to \(|\Psi _{s_i} \rangle = \sqrt{1-s_i}|0\rangle + \sqrt{s_i} | 1 \rangle \). After free time evolution the readout node values are obtained as the ensemble averages of the spins \(\langle Z \rangle \).

Back to article page