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

Local overfitting control via leverages.

Gaétan Monari1, Gérard Dreyfus

  • 1Ecole Supérieure de Physique et de Chimie Industrielles de la Ville de Paris, Laboratoire d'Electronique, F 75005 Paris, France. Gaetan.Monari@sollac.usinor.com

Neural Computation
|May 22, 2002
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

Can Graph Machines Accurately Estimate <sup>13</sup>C NMR Chemical Shifts of Benzenic Compounds?

Molecules (Basel, Switzerland)·2024
Same author

Self-Modulation of Gamma-Band Synchronization through EEG-Neurofeedback Training in the Elderly.

Journal of integrative neuroscience·2024
Same author

Detecting Anosognosia from the Prodromal Stage of Alzheimer's Disease.

Journal of Alzheimer's disease : JAD·2023
Same author

Can a failure in the error-monitoring system explain unawareness of memory deficits in Alzheimer's disease?

Cortex; a journal devoted to the study of the nervous system and behavior·2023
Same author

Probing the posture with machine learning provides physiological evidence supporting the enhanced body awareness hypothesis in trait mindfulness.

Frontiers in physiology·2022
Same author

How machine learning could be used in clinical practice during an epidemic.

Critical care (London, England)·2020
Same journal

A Model-Free Reinforcement Learning Implementation of Decision Making Under Uncertainty by Sequential Sampling.

Neural computation·2026
Same journal

DROP: Distributional and Regular Optimism and Pessimism for Reinforcement Learning.

Neural computation·2026
Same journal

Hierarchical Active Inference Using Successor Representations.

Neural computation·2026
Same journal

W-Kernel and Its Principal Space for Frequentist Evaluation of Bayesian Estimators.

Neural computation·2026
Same journal

A Hidden Markov Model-Inspired Sequence Classification Method for Hyperdimensional Computing.

Neural computation·2026
Same journal

Sparse Graphical Modeling for Electrophysiological Phase-Based Connectivity Using Circular Statistics.

Neural computation·2026
See all related articles

This study introduces a novel method to prevent overfitting in black box models by analyzing sample leverage, or data point influence. This technique aids in selecting optimal nonlinear models based on complexity and training data influence.

Area of Science:

  • Machine Learning
  • Statistical Modeling

Background:

  • Overfitting occurs when models learn training data too specifically, leading to poor generalization.
  • Black box models, common in complex tasks, often face overfitting challenges.
  • Existing model selection methods may not adequately address overfitting in nonlinear models.

Purpose of the Study:

  • To introduce a novel approach for mitigating overfitting in black box models.
  • To develop a model selection methodology based on sample leverage and confidence intervals.
  • To enable selection between models of varying complexities and different minima.

Main Methods:

  • Estimation of sample leverages to quantify data point influence on model parameters.
  • Utilizing confidence intervals in conjunction with leverage estimation.

Related Experiment Videos

  • Developing a selection method applicable to nonlinear models.

    • Main Results:

    • The proposed method effectively identifies and addresses overfitting by considering sample influence.
    • The methodology facilitates model selection across different complexities and cost function minima.
    • Demonstrated capability to select among neural networks with varying numbers of hidden units.

    Conclusions:

    • Sample leverage analysis provides a robust framework for combating overfitting in black box models.
    • The derived methodology offers a comprehensive approach to nonlinear model selection.
    • This work contributes to building more generalizable and reliable machine learning models.