The delta test: The 1-NN estimator as a feature selection criterion

Eirola, Emil; Lendasse, Amaury; Corona, Francesco; Verleysen, Michel

doi:10.1109/ijcnn.2014.6889560

Cited by 8 publications

(3 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(a) The delta test (DT) method: a noise variance estimator based on the concept of nearest neighbors (NNs) (Eirola et al, 2008;Pi et al, 1994). For a given set of input parameters 𝜃 𝑖 (𝑖 = 1, … , 𝑚) and associated output 𝑌, the assumption is that there is a functional dependence between them:…”

Section: Design Of the Parameter Sensitivity Analysismentioning

confidence: 99%

Comparing the Impacts of Single- and Multi-Objective Optimization on the Parameter Estimation and Performance of a Land Surface Model

Xu,

Zhu,

Zhang

et al. 2024

Preprint

View full text Add to dashboard Cite

Abstract. In land surface models (LSMs), precise parameter specification is crucial to reduce the inherent uncertainties and enhance the accuracy of the simulation. However, due to the multi-output nature of LSMs, the impact of different optimization strategies (e.g., single- and multi-objective optimization) on the optimization outcome and efficiency remains ambiguous. In this study, we applied a revised particle evolution Metropolis sequential Monte Carlo (PEM-SMC) algorithm for both single- and multi-objective optimization of the Common Land Model (CoLM), constrained by latent heat flux (LE) and net ecosystem exchange (NEE) measurements from a typical evergreen needle-leaf forest observation site. The results reveal that the revised PEM-SMC algorithm, demonstrates a robust ability to tackle the multi-dimensional parameter optimization challenge for LSMs. The sensitive parameters for different target outputs can exhibit conflicting optimal values, resulting in single-objective optimization improving the simulation performance for a specific objective at the expense of sacrificing the accuracy for other objectives. For instance, solely optimizing for LE reduced the root-mean-square error (RMSE) of the simulated and observed LE by 20 % but increased the RMSE of the NEE by 97 %. Conversely, multi-objective optimization can not only ensure that the optimized parameter values are physically sound but also balances the simulation performance for both LE and NEE, as evidenced by the decrease in RMSE for LE and NEE of 7.2 W/m2 and 0.19 μmol m-2 s-1, respectively. In conclusion, these findings reveal that comprehensively integrating the various available observational data for multi-objective optimization is preferable for parameter calibration in complex models.

show abstract

Section: Design Of the Parameter Sensitivity Analysismentioning

confidence: 99%

Comparing the Impacts of Single- and Multi-Objective Optimization on the Parameter Estimation and Performance of a Land Surface Model

Xu,

Zhu,

Zhang

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…The MGA (multi-objective genetic algorithm) [55] is a new revision of the classical genetic algorithm involving a multi-objective selection operator. This operator is designed to determine two different individuals to cross and evaluate, one individual associated with the MI dimension [56] and the other individual selected when taking into account its score in the Delta Test [57,58]. In the resulting crossover, the individuals are selected according to the two different criteria, thus, the algorithms performs multi-objective optimization.…”

Section: Development Of the Measurement Scalesmentioning

confidence: 99%

Assessing university students' perception of academic quality using machine learning

Perales

Liébana‐Cabanillas

Sánchez‐Fernández

et al. 2020

ACI

View full text Add to dashboard Cite

PurposeThe aim of this research is to assess the influence of the underlying service quality variable, usually related to university students' perception of the educational experience. Another aspect analysed in this work is the development of a procedure to determine which variables are more significant to assess students' satisfaction.Design/methodology/approachIn order to achieve both goals, a twofold methodology was approached. In the first phase of research, an assessment of the service quality was performed with data gathered from 580 students in a process involving the adaptation of the SERVQUAL scale through a multi-objective optimization methodology. In the second phase of research, results obtained from students were compared with those obtained from the teaching staff at the university.FindingsResults from the analysis revealed the most significant service quality dimensions from the students' viewpoint according to the scores that they provided. Comparison of the results with the teaching staff showed noticeable differences when assessing academic quality.Originality/valueSignificant conclusions can be drawn from the theoretical review of the empirical evidences obtained through this study helping with the practical design and implementation of quality strategies in higher education especially in regard to university education.

show abstract

“…Among these subsets of features, the one minimising the value of δ with Y will be selected. The Delta Test has also been widely used for feature selection [21,22].…”

Section: Feature Selection With Noise Variancementioning

confidence: 99%

Smoothness Bias in Relevance Estimators for Feature Selection in Regression

Degeest

Verleysen

Frénay‬

2018

IFIP Advances in Information and Communication Technology

Self Cite

View full text Add to dashboard Cite

Selecting features from high-dimensional datasets is an important problem in machine learning. This paper shows that in the context of filter methods for feature selection, the estimator of the criterion used to select features plays an important role; in particular the estimators may suffer from a bias when comparing smooth and non-smooth features. This paper analyses the origin of such bias and investigates whether this bias influences the results of the feature selection process. Results show that non-smooth features tend to be penalised especially in small datasets.

show abstract

The delta test: The 1-NN estimator as a feature selection criterion

Cited by 8 publications

References 33 publications

Comparing the Impacts of Single- and Multi-Objective Optimization on the Parameter Estimation and Performance of a Land Surface Model

Comparing the Impacts of Single- and Multi-Objective Optimization on the Parameter Estimation and Performance of a Land Surface Model

Assessing university students' perception of academic quality using machine learning

Smoothness Bias in Relevance Estimators for Feature Selection in Regression

Contact Info

Product

Resources

About