Evaluation of estimation algorithms part I: incomprehensive measures of performance

Abstract

Practical metrics for performance evaluation of estimation algorithms are discussed. A variety of metrics useful for evaluating various aspects of the performance of an estimation algorithm is introduced and justified. They can be classified in two different ways: 1) absolute error measures (without a reference), relative error measures (with a reference), or frequency counts (of some events), and 2) optimistic (i.e., how good the performance is), pessimistic (i.e., how bad the performance is), or balanced (neither optimistic nor pessimistic). Pros and cons of these metrics and the widely-used RMS error are explained. The paper advocates replacing the RMS error in many cases by a measure called average Euclidean error.Manuscript

Deep Gaussian Process metamodeling of sequentially sampled non-stationary response surfaces

“…This routine was found empirically to be much less prone to local minima as we assist the optimizer by limiting the number of optimizable variables in the first stage. The performance of the models is assessed using three metrics: Root-Mean-Square-Error (RMSE), Root-Relative-Square-Error (RRSE) and the Mean-Square-Error (MSE), which is the square of the RMSE (Li and Zhao 2006):…”

Experimental evaluation of joint position estimation using inertial sensors

“…The rationale for choosing the average Euclidean error over e.g. the more commonly used root mean square error (RMSE) is that it may be interpreted as the average physical distance between estimates with respect to the true joint center [12]. This results in a more intuitive understanding as to how close the estimators are to estimating the true position of the joint center.…”

A multihypothesis set approach for mobile robot localization using heterogeneous measurements provided by the Internet of Things

“…Localization accuracy might be assessed according to average or instantaneous error criteria. The average error as the mean square error RMSE (Root Mean Squared Error) reflects the overall performance of the estimator [28]. However it is also interesting to measure the accuracy on a temporal horizon to consider the behavior of the estimator in borderline cases.…”