Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Videos

Partial retraining: a new approach to input relevance determination.

P van de Laar¹, T Heskes, S Gielen

¹Department of Medical Physics and Biophysics, University of Nijmegen, The Netherlands. pierre@mbfys.kun.nl

International Journal of Neural Systems

|July 13, 1999

Summary

We introduce partial retraining, a novel algorithm for assessing input variable relevance in trained neural networks. This method demonstrates superior performance compared to existing techniques in numerical experiments.

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

Hidradenitis suppurativa is associated with childhood and lifetime traumatic events: a case-control study.

Journal of the European Academy of Dermatology and Venereology : JEADV·2020

Same author

[Bed capacity analysis for an intensive care unit : Retrospective analysis and projection for need of intensive care beds in over 65 year olds with selected cardiovascular diseases].

Medizinische Klinik, Intensivmedizin und Notfallmedizin·2020

Same author

Exercise-based cardiac rehabilitation in patients with reduced left ventricular ejection fraction: The Cardiac Rehabilitation Outcome Study in Heart Failure (CROS-HF): A systematic review and meta-analysis.

European journal of preventive cardiology·2019

Same author

[Acute iron poisoning : Successful treatment of a potentially fatal suicidal attempt].

Medizinische Klinik, Intensivmedizin und Notfallmedizin·2018

Same author

Role of conduct problems in the relation between Attention-Deficit Hyperactivity disorder, substance use, and gaming.

European neuropsychopharmacology : the journal of the European College of Neuropsychopharmacology·2018

Same author

Lung protective ventilation and hospital survival of cardiac intensive care patients.

Medizinische Klinik, Intensivmedizin und Notfallmedizin·2015

Area of Science:

Machine Learning
Artificial Intelligence
Computational Neuroscience

Background:

Determining input variable relevance is crucial for understanding and optimizing trained neural networks.
Existing methods like constant substitution, weight magnitude analysis, and Optimal Brain Surgeon have limitations.

Purpose of the Study:

To introduce and evaluate a new algorithm, partial retraining, for assessing input variable relevance in neural networks.
To compare the efficacy of partial retraining against established relevance determination techniques.

Main Methods:

The partial retraining algorithm is proposed as a novel approach to feature relevance.
Comparative numerical experiments were conducted on both artificial and real-world datasets.

Related Experiment Videos

Main Results:

Partial retraining consistently outperformed competing methods in relevance determination tasks.
The algorithm proved effective across diverse problem types, including artificial and real-world scenarios.

Conclusions:

Partial retraining offers a more effective approach to identifying relevant input variables in neural networks.
This algorithm provides a valuable tool for model interpretation and feature selection in machine learning.