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

Learning with incomplete information and the mathematical structure behind it.

Reimer Kühn1, Ion-Olimpiu Stamatescu

  • 1Department of Mathematics, King's College, London, UK. reimer.kuehn@kcl.ac.uk

Biological Cybernetics
|May 31, 2007
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

Gaussian level-set percolation on complex networks.

Physical review. E·2024
Same author

Level-set percolation of Gaussian random fields on complex networks.

Physical review. E·2024
Same author

Distribution of the number of cycles in directed and undirected random regular graphs of degree 2.

Physical review. E·2023
Same author

Conformational Heterogeneity and Interchain Percolation Revealed in an Amorphous Conjugated Polymer.

ACS nano·2022
Same author

Overcoming the complexity barrier of the dynamic message-passing method in networks with fat-tailed degree distributions.

Physical review. E·2021
Same author

Fate of articulation points and bredges in percolation.

Physical review. E·2021
Same journal

Harmonic memory in phasor neural networks.

Biological cybernetics·2026
Same journal

Correction: Decreased spinal inhibition leads to undiversified locomotor patterns.

Biological cybernetics·2026
Same journal

Foundational issues of network models in biology.

Biological cybernetics·2026
Same journal

Dynamical mechanisms for coordinating long-term working memory based on the precision of spike-timing in cortical neurons.

Biological cybernetics·2026
Same journal

Distinct dopaminergic spike-timing-dependent plasticity rules are suited to different functional roles.

Biological cybernetics·2026
Same journal

Fluctuation-response relations for a two-stage population of spiking neurons stimulated by common noise.

Biological cybernetics·2026
See all related articles

This study introduces a two-phase learning model for incomplete information, combining Hebbian learning with reinforcement-based unlearning. Optimal learning and generalization require a specific ratio of learning rates, crucial for tasks like robot navigation.

Area of Science:

  • Machine Learning
  • Computational Neuroscience
  • Artificial Intelligence

Background:

  • Learning with incomplete information, such as delayed reinforcement, poses significant challenges in artificial systems.
  • Traditional learning models often struggle with scenarios where feedback is not immediate or globally assessed.

Purpose of the Study:

  • To investigate a novel two-phase learning model incorporating Hebbian associative learning and reinforcement-driven unlearning.
  • To analyze the impact of the learning rate ratio on learning efficiency and generalization capabilities.

Main Methods:

  • Simulations and analytical studies within a student-teacher framework.
  • Analysis applied to single-layer networks and committee machines.
  • Investigated the influence of the ratio (lambda) between Hebbian learning and unlearning rates.

Related Experiment Videos

Main Results:

  • Asymptotically perfect generalization is achieved only when the learning rate ratio (lambda) exceeds a critical value (lambda_c).
  • Generalization error decays as a power law with the number of examples, dependent on lambda.
  • The model's key features demonstrate robustness against variations in microscopic details.

Conclusions:

  • The proposed two-phase learning model offers a viable approach for handling learning with delayed or incomplete information.
  • The critical learning rate ratio is essential for achieving high generalization performance.
  • The model's principles are applicable to real-world problems, including robot navigation and stimulus identification.