Online algorithm based on support vectors for orthogonal regression

Applying machine learning to accelerate new materials development

“…而机器学习方法在大数据的支持下可 以发挥更大的作用 [59][60][61] , 因此, 将机器学习与现有的理 论计算方法结合起来, 可以加速理论模拟工作的进展. 将密度泛函理论与机器学习结合起来的方法有很 多, 最简单, 最直接的方法就是利用密度泛函理论的计 算结果作为机器学习的训练数据进行学习与预测 [62] , 这种思想已有一些实例, Ceder实验组 [ [61] , 同时, 机器学习方法里也包 含增量学习 [47,74] 和在线学习 [75][76][77] [81] . 包括:…”

Real-time sufficient dimension reduction through principal least squares support vector machines

“…In the early 1970s, real-time singular value decomposition was discussed in [1], and quick updates of linear regression was introduced in [2]. Recent developments of real-time regression include fast robust linear regression [3], real-time semiparametric regression [4], online orthogonal regression [5], and fast regression for large data [6]. In terms of applications, real-time algorithms have been applied to head pose estimation [7], stock market analysis [8], and human action recognition [9].…”

A generalized power iteration method for solving quadratic problem on the Stiefel manifold

“…Recently, due to the wide applications of the orthogonal regression in computer science, see [5], [6], solving the unbalanced OPP (UOPP) is under increasing concern. Multiple approaches are proposed to solve UOPP such as the expansion balanced algorithm (EB), the right hand side and the left hand side relaxation (RSR), (LSR), the successive projection (SP) and the Lagrangian relaxation (LR).…”