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

Confidence estimation methods for neural networks: a practical comparison.

G Papadopoulos1, P J Edwards, A F Murray

  • 1Department of Electronics and Electrical Engineering, University of Edinburgh, Edinburgh EH9 3JL, UK.

IEEE Transactions on Neural Networks
|February 6, 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

Growth and population structure of Lodoicea maldivica in natural stands in Seychelles.

Plant biology (Stuttgart, Germany)·2024
Same author

First Constraints from DAMIC-M on Sub-GeV Dark-Matter Particles Interacting with Electrons.

Physical review letters·2023
Same author

Changes in germinability, ABA content and ABA embryonic sensitivity in developing seeds of Sorghum bicolor (L.) Moench. induced by water stress during grain filling.

The New phytologist·2021
Same author

B3 agonists or anticholinergics in the treatment of the lower urinary tract dysfunction in patients with multiple sclerosis?-A randomized study.

World journal of urology·2021
Same author

Particulate emissions from L-Category vehicles towards Euro 5.

Environmental research·2019
Same author

Three- versus six-month adjuvant FOLFOX or CAPOX for high-risk stage II and stage III colon cancer patients: the efficacy results of Hellenic Oncology Research Group (HORG) participation to the International Duration Evaluation of Adjuvant Chemotherapy (IDEA) project.

Annals of oncology : official journal of the European Society for Medical Oncology·2019
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

Estimating prediction confidence in feedforward neural networks is crucial. The approximate Bayesian method offers superior global performance for tasks like predicting paper curl, outperforming maximum likelihood and bootstrap techniques.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Statistical Modeling

Background:

  • Feedforward neural networks, including multilayer perceptrons, are extensively utilized for regression and classification.
  • Accurate prediction confidence estimation is vital for reliable decision-making in these applications.
  • Uncertainty arises from both inherent data noise and potential model parameter misspecification.

Purpose of the Study:

  • To present and compare three distinct methods for estimating prediction confidence in neural networks.
  • To evaluate these methods considering uncertainties from data noise and model misspecification.
  • To assess performance using both artificial datasets and a real-world industrial regression problem (paper curl prediction).

Main Methods:

  • Maximum Likelihood Estimation

Related Experiment Videos

  • Approximate Bayesian Inference
  • Bootstrap Resampling Technique

    • Main Results:

    • The approximate Bayesian approach demonstrated superior global performance in confidence estimation.
    • Treating data noise variance as input-dependent was found appropriate for the paper curl prediction task.
    • Mean coverage probability assesses global performance, while standard deviation is unreliable for local performance evaluation.

    Conclusions:

    • The approximate Bayesian method provides a more reliable approach to prediction confidence estimation in feedforward neural networks.
    • Understanding the limitations of performance metrics (mean vs. standard deviation of coverage) is essential for accurate assessment.
    • Input-dependent noise variance modeling is a key consideration for specific regression tasks.