Specifically, we consider finding the largest subset of data items that can be represented by a linear model to a given accuracy. In this paper we present a sparse robust regression method, termed slise ( Sparse Linear Subset Explanations), that achieves the highest possible theoretical robustness and outperforms many existing state-of-the-art robust regression methods in terms of scalability on large datasets. Furthermore, robust regression can be used to search for outliers by investigating the data items that do not adhere to the robust model. Linear regression is also often used as a part of other machine learning or data mining algorithms, e.g., in explainable artificial intelligence (Ribeiro et al. ![]() Robust regression methods can be used as almost drop-in replacements for linear regression, which is still widely used because of the inherent interpretability and simplicity. It is, hence, important to consider robust methods that effectively avoid the influence of outliers. This is observed, for instance, in ordinary least-squares ( ols) regression where already a single outlier may lead to arbitrarily large errors (Donoho and Huber 1983). Such items are problematic, since they may negatively influence modelling of the data. ![]() In practically all analyses of real-world data we encounter outliers, i.e., data items that do not follow the same patterns as the majority of the data.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |