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

Neural networks convergence using physicochemical data.

Mati Karelson1, Dimitar A Dobchev, Oleksandr V Kulshyn

  • 1Department of Chemistry, University of Tartu, 2 Jakobi Street, Tartu 51014, Estonia. mati.karelson@ttu.ee

Journal of Chemical Information and Modeling
|September 26, 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

Inhibition of ALKBH5 demethylase of m<sup>6</sup>A pathway potentiates HIV-1 reactivation from latency.

Virology journal·2025
Same author

Dengue Virus Inhibitors as Potential Broad-Spectrum Flavivirus Inhibitors.

Pharmaceuticals (Basel, Switzerland)·2025
Same author

USP14 is crucial for proteostasis regulation and α-synuclein degradation in human SH-SY5Y dopaminergic cells.

Heliyon·2025
Same author

Computational Methods for Modeling Lipid-Mediated Active Pharmaceutical Ingredient Delivery.

Molecular pharmaceutics·2025
Same author

RNA m<sup>6</sup>A methyltransferase activator affects anxiety-related behaviours, monoamines and striatal gene expression in the rat.

Acta neuropsychiatrica·2024
Same author

Epitranscriptomic regulation in fasting hearts: implications for cardiac health.

RNA biology·2024
Same journal

QSAR in the Browser: An Interactive Cheminformatics Web Application.

Journal of chemical information and modeling·2026
Same journal

FoldDoF: Utilizing the Primary Degrees of Freedom of Protein Backbone for Geometric Modeling and Generation.

Journal of chemical information and modeling·2026
Same journal

Derisking Affinity Optimization for Macrocycles and Cyclic Peptides: High-Precision Free Energy Simulations across Five Diverse Targets.

Journal of chemical information and modeling·2026
Same journal

An End-User Audit of Reproducibility, Data Leakage, and Overfitting of the Top-Ranked ADMET Prediction Models in TDC Leaderboards.

Journal of chemical information and modeling·2026
Same journal

PFASGroups: An Open-Source Framework for Automated Identification, Structural Classification, and Prioritization of Per- and Polyfluoroalkyl Substances.

Journal of chemical information and modeling·2026
Same journal

DeepKbhb: Context-Aware Prediction of Human Lysine β-Hydroxybutyrylation Sites.

Journal of chemical information and modeling·2026
See all related articles

The Levenberg-Marquardt algorithm, used in back-propagation neural networks, converges faster and offers superior prediction compared to conjugate gradient and delta rule optimizers for QSAR/QSPR modeling.

Area of Science:

  • Computational Chemistry
  • Cheminformatics
  • Machine Learning

Background:

  • Quantitative Structure-Activity Relationship (QSAR) and Quantitative Structure-Property Relationship (QSPR) models are crucial for predicting chemical compound properties.
  • Neural networks are powerful tools for QSAR/QSPR modeling, but their performance depends heavily on the chosen optimization algorithm.
  • Comparing optimization algorithms is essential for enhancing the accuracy and efficiency of neural network-based chemical modeling.

Purpose of the Study:

  • To compare the convergence speed and predictive accuracy of three optimization algorithms: Levenberg-Marquardt, conjugate gradient, and delta rule.
  • To evaluate the performance of these algorithms in the context of neural network-based QSAR/QSPR modeling.
  • To identify the most effective optimization algorithm for physicochemical data analysis.

Related Experiment Videos

Main Methods:

  • Simulated neural networks were constructed using Levenberg-Marquardt, conjugate gradient, and delta rule optimization algorithms.
  • Eight diverse physicochemical data sets were utilized, each containing a substantial number of compounds.
  • Model prediction accuracy was assessed using validation sets independent of the training data.

Main Results:

  • The Levenberg-Marquardt algorithm demonstrated faster convergence compared to the conjugate gradient and delta rule algorithms.
  • Neural networks optimized with Levenberg-Marquardt generally provided better predictive performance across the tested data sets.
  • Analysis of weight changes and distribution revealed the superior functional dependence of the Levenberg-Marquardt algorithm.

Conclusions:

  • The Levenberg-Marquardt algorithm is a superior choice for optimizing neural networks in QSAR/QSPR modeling due to its faster convergence and enhanced predictive capabilities.
  • The findings provide valuable insights for researchers seeking to improve the accuracy of predictive models in cheminformatics and computational chemistry.
  • This study highlights the importance of algorithm selection in achieving reliable and efficient QSAR/QSPR predictions.