Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Performing feature selection with multilayer perceptrons.

Enrique Romero1, Josep María Sopena

  • 1Departament de Llenguatges i Sistemes Informàtics, Universitat Politècnica de Catalunya, 08034 Barcelona, Spain. eromero@lsi.upc.edu

IEEE Transactions on Neural Networks
|March 13, 2008
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

The fascinating complexity of seagrass bio-fibres: insights from bio-chemo-hygro mechanical analysis for their reuse as soil reinforcement.

Scientific reports·2026
Same author

Machine Learning Prediction of Cardiovascular Risk in Type 1 Diabetes Mellitus Using Radiomic Features from Multimodal Retinal Images.

Ophthalmology science·2025
Same author

Mean Field Initialization of the Annealed Importance Sampling Algorithm for an Efficient Evaluation of the Partition Function Using Restricted Boltzmann Machines.

Entropy (Basel, Switzerland)·2025
Same author

Radiomics-Based Assessment of OCT Angiography Images for Diabetic Retinopathy Diagnosis.

Ophthalmology science·2022
Same author

On-The-Fly Syntheziser Programming with Fuzzy Rule Learning.

Entropy (Basel, Switzerland)·2020
Same author

Unraveling response to temozolomide in preclinical GL261 glioblastoma with MRI/MRSI using radiomics and signal source extraction.

Scientific reports·2020
Same journal

Universal perceptron and DNA-like learning algorithm for binary neural networks: LSBF and PBF implementations.

IEEE transactions on neural networks·2013
Same journal

Guest editorial: special section on white box nonlinear prediction models.

IEEE transactions on neural networks·2011
Same journal

Data-based fault-tolerant control of high-speed trains with traction/braking notch nonlinearities and actuator failures.

IEEE transactions on neural networks·2011
Same journal

Guest editorial: special section on data-based control, modeling, and optimization.

IEEE transactions on neural networks·2011
Same journal

Neural network-based multiple robot simultaneous localization and mapping.

IEEE transactions on neural networks·2011
Same journal

Data-driven model-free adaptive control for a class of MIMO nonlinear discrete-time systems.

IEEE transactions on neural networks·2011
See all related articles

Retraining networks during feature selection significantly improves performance, even with higher computational costs. Nonintuitive findings suggest overtraining can be as effective as early stopping for better results.

Area of Science:

  • Machine Learning
  • Computational Intelligence
  • Data Science

Background:

  • Wrapper feature selection (FS) methods are crucial for optimizing model performance.
  • Multilayer perceptrons (MLPs) are widely used in classification tasks.
  • Sequential Backward Selection (SBS) is a common FS algorithm.

Purpose of the Study:

  • To investigate the impact of two key decision parameters in wrapper FS using MLPs and SBS.
  • To evaluate the effect of network retraining and stopping criteria on FS performance.

Main Methods:

  • An experimental study was conducted using wrapper feature selection with MLPs.
  • The Sequential Backward Selection (SBS) procedure was employed.
  • Two decision issues were analyzed: the stopping criterion and network retraining before saliency computation.

Related Experiment Videos

Main Results:

  • Increasing computational cost by retraining the network with each removed feature significantly improved performance.
  • A nonintuitive finding indicated that forcing overtraining can be as beneficial as early stopping.
  • Overall results showed significant improvement compared to using the entire feature set.

Conclusions:

  • Network retraining in wrapper FS, despite higher costs, yields substantial performance gains.
  • The choice of stopping criterion, including overtraining, impacts FS effectiveness.
  • Feature selection using this enhanced SBS approach with MLPs outperforms using all variables.