A new algorithm for support vector regression with automatic selection of hyperparameters

The hyperparameters in support vector regression (SVR) determine the effectiveness of the support vectors with fitting and predictions. However, the choice of these hyperparameters has always been challenging in both theory and practice. The v-support vector regression eliminates the need to specify an is an element of value elegantly, but at the cost of specifying or postulating a v value. We propose an extended primal objective function arising from probability regularization leading to an automatic selection of is an element of, and we can express v as an explicit function of is an element of. The resultant hyperparameter values can be interpreted as 'working' values required only in training but not testing or prediction. This regularized algorithm, namely is an element of*-SVR, automatically provides a data-dependent is an element of and is found to have a close connection to the v-support vector regression in the sense that v as a fraction is a sensible function of is an element of. The is an element of*- SVR automatically selects both v and is an element of values. We illustrate these findings with some public benchmark datasets.(C) 2022 Elsevier Ltd. All rights reserved.

Automated summary

The choice of hyperparameters in support vector regression has always been challenging. This paper proposes an extended primal objective function based on probability regularization, which automatically selects appropriate parameters and has a close connection to v-support vector regression.