Definition
A supervised learning method that constructs decision boundaries using a margin criterion and can handle nonlinear data through kernel methods.
In their 1995 paper, Cortes and Vapnik extended support-vector networks to nonseparable training data and presented classification experiments. Research on separable cases already existed, so the method’s entire history cannot be compressed into a single year.
Support vector machines are an important branch of statistical learning and need not be modeled as a historical successor to neural networks.
Principles and impact
Problems solved
- Constructed classification boundaries with generalization capability from finite samples.
Principles
- Balance the classification margin against training error, and express nonlinear boundaries through feature mappings or kernel functions.
Improvements
Not recorded
Significance
- Represented an important supervised learning approach independent of deep neural networks.
History and applications
People
Corinna Cortes、Vladimir Vapnik
Organizations
Not recorded
Events
Not recorded
Applications
Classification、Pattern recognition
Limitations
- Performance depends on the choice of features, kernels, and parameters, and training on large samples may be costly.
Upstream and downstream relations
Upstream
Sources: Support-vector Networks
Downstream
No direct downstream relations recorded