Quantum Algorithms for Prediction Based on Ridge Regression

TL;DR

This paper introduces a quantum algorithm for ridge regression that efficiently computes optimal parameters and regularization hyperparameters, overcoming classical limitations and offering exponential speedup, applicable to non-sparse matrices.

Contribution

It presents a novel quantum ridge regression algorithm that improves efficiency, handles non-sparse matrices, and can serve as a subroutine in broader quantum machine learning applications.

Findings

Achieves exponential speedup over classical methods.

Handles non-sparse matrices unlike some existing quantum algorithms.

Overcomes multicollinearity and overfitting issues.

Abstract

We propose a quantum algorithm based on ridge regression model, which get the optimal fitting parameters w and a regularization hyperparameter {\alpha} by analysing the training dataset. The algorithm consists of two subalgorithms. One is generating predictive value for a new input, the way is to apply the phase estimation algorithm to the initial state |Xi and apply the controlled rotation to the eigenvalue register. The other is finding an optimal regularization hyperparameter {\alpha} , the way is to apply the phase estimation algorithm to the initial state |yi and apply the controlled rotation to the eigenvalue register. The second subalgorithm can compute the whole training dataset in parallel that improve the efficiency. Compared with the classical ridge regression algorithm, our algorithm overcome multicollinearity and overfitting. Moreover, it have exponentially faster. What's more, our algorithm can deal with the non-sparse matrices in comparison to some existing quantum algorithms and have slightly speedup than the existing quantum counterpart. At present, the quantum algorithm has a wide range of application and the proposed algorithm can be used as a subroutine of other quantum algorithms.