An Improved Hybrid Conjugate Gradient Method With Spectral Strategy and Its Applications in Covid-19 Prediction

This paper introduces a hybrid conjugate gradient (CG) method for unconstrained optimization with a spectral strategy, inspired by key advancements in existing CG techniques. The proposed method guarantees a descent direction at every iteration, regardless of the line search scheme employed. Its global convergence is rigorously established under the Wolfe line search conditions. Numerical experiments on benchmark optimization problems demonstrate that the proposed method outperforms the FR and RMIL methods across multiple performance metrics. Furthermore, its effectiveness is showcased in a neural network framework for predicting chickenpox and COVID-19 infection cases, highlighting its practical applicability in real-world scenarios.