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

Safe semi-supervised learning based on weighted likelihood.

Masanori Kawakita1, Jun'ichi Takeuchi1

  • 1Graduate School of Information Science and Electrical Engineering, Kyushu University, 744 Motooka, Nishi-Ku, Fukuoka, 819-0395, Japan.

Neural Networks : the Official Journal of the International Neural Network Society
|March 18, 2014
PubMed
Summary

This study introduces a safer semi-supervised learning approach that guarantees performance no worse than supervised learning. By modifying weights, the method ensures safety across broader conditions, enhancing machine learning reliability.

Related Concept Videos

Weighted Mean00:57

Weighted Mean

5.5K
While taking the arithmetic, geometric, or harmonic mean of a sample data set, equal importance is assigned to all the data points. However, all the values may not always be equally important in some data sets. An intrinsic bias might make it more important to give more weightage to specific values over others.
For example, consider the number of goals scored in the matches of a tournament. While computing the average number of goals scored in the tournament, it may be more important to...
5.5K
Survival Tree01:19

Survival Tree

499
Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
 Building a Survival Tree
Constructing a...
499

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Approximate spectral decomposition of Fisher information matrix for simple ReLU networks.

Neural networks : the official journal of the International Neural Network Society·2023
Same author

A simple optimization can improve the performance of single feature polymorphism detection by Affymetrix expression arrays.

BMC genomics·2010
Same author

Common peak approach using mass spectrometry data sets for predicting the effects of anticancer drugs on breast cancer.

Cancer informatics·2009
Same author

Boosting method for local learning in statistical pattern recognition.

Neural computation·2008

Area of Science:

  • Machine Learning
  • Statistical Learning Theory

Background:

  • Semi-supervised learning utilizes labeled and unlabeled data but often requires restrictive assumptions.
  • Prior methods may underperform supervised learning if assumptions are violated.
  • A prior safe semi-supervised method (Sokolovska et al., 2008) had limited applicability.

Purpose of the Study:

  • To develop a universally safe semi-supervised learning method.
  • To extend the safety guarantees of existing methods to wider scenarios.
  • To provide theoretical and empirical validation of the proposed approach.

Main Methods:

  • Modification of the weighting scheme in a weighted likelihood framework.
  • Theoretical analysis to prove safety guarantees under relaxed assumptions.
Keywords:
Density ratioEstimating functionGeometrical interpretationSemi-supervised learningStatistical paradoxWeighted likelihood

Related Experiment Videos

  • Geometric interpretation of safety proofs linked to statistical paradoxes.
  • Numerical experiments to evaluate performance.

    • Main Results:

    • The proposed method is proven safe for semi-supervised learning when the number of unlabeled data is greater than or equal to labeled data (n ≤ n').
    • Asymptotic safety is achieved even when unlabeled data is less than labeled data (n' < n) through weight modification.
    • The method demonstrates practical effectiveness in numerical experiments.

    Conclusions:

    • The enhanced semi-supervised learning approach offers robust safety guarantees across diverse conditions.
    • This work significantly broadens the applicability of safe semi-supervised learning.
    • The findings contribute to the theoretical understanding of learning with limited labeled data.