Learning Regression Models with Guaranteed Error Bounds
Clemens Otte · 2015
Abstract. The combination of a symbolic regression model with a resid-ual Gaussian Process is proposed for providing an interpretable model with improved accuracy. While the learned symbolic model is highly in-terpretable the residual model usually is not. However, by limiting the output of the residual model to a defined range a worst-case guarantee can be given in the sense that the maximal deviation from the symbolic model is always below a defined limit. When ranking the accuracy and in-terpretability of several different approaches on the SARCOS data bench-mark the proposed combination yields the best result. 1