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

Weighted quality estimates in machine learning.

Levon Budagyan1, Ruben Abagyan

  • 1Molsoft LLC, 3366 North Torrey Pines Court Suite 300, San Diego, CA 92037, USA. levon@molsoft.com

Bioinformatics (Oxford, England)
|August 29, 2006
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

Pharmacological proximities in the GPCR family discovered using contact-informed amino-acid and binding pocket similarities.

bioRxiv : the preprint server for biology·2026
Same author

A Polypharmacology-Driven Approach to Alzheimer's Disease and Tauopathies: Rational Design, Synthesis and Characterization of Amino-Pyrazole-Based Multikinase (GSK-3β/FYN-α/DYRK1A) Inhibitors.

Journal of medicinal chemistry·2026
Same author

Menthol's disruptive effects on kanamycin-resistant Escherichia coli energy metabolism and ion fluxes.

Biophysical reports·2025
Same author

Structure based design, synthesis and identification of novel covalent reversible dual TLR2/TLR9 small molecule antagonists.

Bioorganic & medicinal chemistry letters·2025
Same author

Identifying Allosteric Small-Molecule Binding Sites of Inactive NS2B-NS3 Proteases of Pathogenic <i>Flaviviridae</i>.

Viruses·2025
Same author

Co-Occurring Infections in Cancer Patients Treated with Checkpoint Inhibitors Significantly Increase the Risk of Immune-Related Adverse Events.

Cancers·2024

This study introduces inverse density weighting to improve machine learning model quality assessment. This method provides more realistic estimates of model generalization by addressing biases in training data.

Area of Science:

  • Computational Biology
  • Machine Learning
  • Bioinformatics

Background:

  • Machine learning model quality assessment is often overly optimistic due to biased training datasets.
  • Current methods reduce bias by filtering data, which can discard valuable information.

Purpose of the Study:

  • To develop a novel approach for calculating prediction model quality that accounts for data distribution.
  • To improve the accuracy and reliability of machine learning model evaluation.

Main Methods:

  • Introduced inverse density weighting based on a postulated distance metric.
  • Reformulated the Vapnik-Chervonenkis theorem to derive space-uniform error estimates.
  • Applied density-weighted quality estimates to a signal peptide prediction task.

Related Experiment Videos

Main Results:

  • Demonstrated that inverse density weighting provides more realistic estimates of model generalization.
  • Showcased improved performance on a biased dataset compared to unweighted cross-validation.
  • Successfully developed a Support Vector Machine (SVM) model using the full dataset for signal peptide prediction.

Conclusions:

  • Inverse density weighting offers a more robust method for evaluating machine learning models.
  • This approach enhances model generalization assessment and overcomes limitations of traditional data filtering.
  • The method has practical applications in bioinformatics, such as signal peptide prediction.