Behavior prediction of fiber optic temperature sensor based on hybrid classical quantum regression model

In this research work, a quantum regression model (QRM) is proposed by combining an autoencoder and a dressed quantum circuit (DQC) to predict the behavior of fiber optic temperature sensors. As the experimental data gathered during our observations was limited to effectively train the proposed QRM model, we employed an autoencoder to expand the dataset. We examined the regression performance of the QRM by running multiple simulations by varying the quantum hyperparameters such as quantum depth \(\varvec{Q_{depth}}\) ${\mathit{Q}}_{\mathit{depth}}$ , number of shots \(\varvec{n_{shots}}\) ${\mathit{n}}_{\mathit{shots}}$ , and the number of qubits \(\varvec{n_{qubits}}\) ${\mathit{n}}_{\mathit{qubits}}$ of the quantum node. Moreover, the regression performance with the unknown data exhibits high R-squared \(\varvec{(r^2)}\) $\mathbf{(}{\mathit{r}}^{\mathbf{2}}\mathbf{)}$ as 0.965, high explained variance \(\varvec{(ExpVar)}\) $\mathbf{(}\mathit{E}\mathit{x}\mathit{p}\mathit{V}\mathit{a}\mathit{r}\mathbf{)}$ as 0.969, and small maximum error \(\varvec{(MaxErr)}\) $\mathbf{(}\mathit{M}\mathit{a}\mathit{x}\mathit{E}\mathit{r}\mathit{r}\mathbf{)}$ as 0.212 for 4 \(\varvec{Q_{depth}}\) ${\mathit{Q}}_{\mathit{depth}}$ , 1500 \(\varvec{n_{shots}}\) ${\mathit{n}}_{\mathit{shots}}$ and 4 \(\varvec{n_{qubits}}\) ${\mathit{n}}_{\mathit{qubits}}$ . Additionally, we proved the superiority performance of the proposed QRM for predicting relative power as it is compared with four conventional machine learning regressors, namely artificial neural network (ANN) regressor, support vector regressor (SVR), decision tree (DT) regressor, and random forest (RF) regressor.