Partial Least Squares Regression on Riemannian Manifolds and Its Application in Classifications
This work addresses a methodological bottleneck for researchers in machine learning and pattern recognition by improving PLSR optimization, though it appears incremental as it builds on existing PLSR techniques.
The authors tackled the problem of suboptimal solutions in partial least squares regression (PLSR) by proposing novel PLSR models on Riemannian manifolds, developing optimization algorithms based on Riemannian geometry to calculate all factors globally, and demonstrated benefits in pattern recognition and object classification experiments.
Partial least squares regression (PLSR) has been a popular technique to explore the linear relationship between two datasets. However, most of algorithm implementations of PLSR may only achieve a suboptimal solution through an optimization on the Euclidean space. In this paper, we propose several novel PLSR models on Riemannian manifolds and develop optimization algorithms based on Riemannian geometry of manifolds. This algorithm can calculate all the factors of PLSR globally to avoid suboptimal solutions. In a number of experiments, we have demonstrated the benefits of applying the proposed model and algorithm to a variety of learning tasks in pattern recognition and object classification.