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

Good practice in large-scale learning for image classification.

Zeynep Akata1, Florent Perronnin2, Zaid Harchaoui3

  • 1Xerox Research Centre Europe, Meylan and INRIA Rhone-Alpes, Montbonnot.

IEEE Transactions on Pattern Analysis and Machine Intelligence
|January 25, 2014
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

Aggregates Classification01:29

Aggregates Classification

1.0K
Aggregate classification is generally based on its size, petrographic characteristics, weight, and source. Size classification ranges from coarse to fine aggregates, defined by the size of the particles. Coarse aggregates are particles that do not pass through ASTM sieve No. 4, and aggregates that pass through the sieve are fine aggregates.
Petrographic classification groups aggregates based on common mineralogical characteristics. Some of the common mineral groups found in aggregates are...
1.0K

You might also read

Related Articles

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

Sort by
Same author

Evaluating large language models for diagnostic reasoning from unstructured clinical narratives in epilepsy.

Communications medicine·2026
Same author

Generative AI as a tool to accelerate the field of ecology.

Nature ecology & evolution·2025
Same author

How should the advancement of large language models affect the practice of science?

Proceedings of the National Academy of Sciences of the United States of America·2025
Same author

The influence of intentions on dream content.

Sleep advances : a journal of the Sleep Research Society·2025
Same author

CoVR-2: Automatic Data Construction for Composed Video Retrieval.

IEEE transactions on pattern analysis and machine intelligence·2024
Same author

Relational Proxies: Fine-Grained Relationships as Zero-Shot Discriminators.

IEEE transactions on pattern analysis and machine intelligence·2024
Same journal

HardFlow: Hard-Constrained Sampling for Flow-Matching Models Via Trajectory Optimization.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Industrial Brain: Self-Evolving Neuro-Symbolic Autonomy with Causal Resilience for Cyber-Physical Systems.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Adaptive Hardness-Driven Dictionary Distillation for Incomplete Streaming View Clustering.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Mixture of Global and Local Experts with Diffusion Transformer for Controllable Face Generation.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Achieving Text-based Person Retrieval with Any Granularity.

IEEE transactions on pattern analysis and machine intelligence·2026
See all related articles

We evaluated Support Vector Machine (SVM) objective functions for large-scale image classification. Online methods offer faster training than batch methods without sacrificing accuracy, improving ImageNet performance by 2.4%.

Area of Science:

  • Computer Science
  • Machine Learning
  • Artificial Intelligence

Background:

  • Large-scale image classification presents significant computational challenges.
  • Support Vector Machines (SVMs) are a powerful tool for classification tasks.
  • Benchmarking different SVM objective functions is crucial for optimizing performance.

Purpose of the Study:

  • To benchmark various SVM objective functions for large-scale image classification.
  • To compare the efficiency and accuracy of online versus batch training methods.
  • To identify best practices for improving SVM performance in image recognition.

Main Methods:

  • Comparison of one-versus-rest, multiclass, ranking, and weighted approximate ranking SVMs.
  • Evaluation of online and batch optimization methods using stochastic gradient descent.

Related Experiment Videos

  • Analysis of training speed and classification accuracy on large datasets (ImageNet).

    • Main Results:

    • Online SVM training methods achieve comparable accuracy to batch methods with significantly faster training speeds.
    • Ranking-based SVMs did not outperform the one-versus-rest strategy on large datasets.
    • Optimal rebalancing of positive/negative examples and early stopping improved one-versus-rest accuracy.
    • State-of-the-art Top-1 accuracy on ImageNet was improved from 16.7% to 19.1%.

    Conclusions:

    • Online SVM training is a scalable and efficient approach for large-scale image classification.
    • The one-versus-rest strategy, when optimized, remains highly competitive.
    • Implementing "good practices" like data rebalancing and early stopping significantly enhances classification performance.