After Reading the paper [1] (you can also find it from url https://www.aaai.org/Papers/AAAI/2005/AAAI05-143.pdf), I take the following notes.
Welcome to point out anything inappropriate.
Xu(2004) develop unsupervised and semi-supervised training techniques for SVMs based on semidefinite programming (SDP). The contribution of this paper is to extend the above work to multi-class SVMs.
做法: find a labeling that would yield an optimal SVM classifier on the resulting training data
优点: the previous work has concentrated on the two-class case, this paper presents a general, multi-class formulation that can be applied to a wider range of natural data sets.
缺点: computationally intensive to produce high quality generalization results
1. Simple introduction about SDP:
SDP is a recent extension of linear and quadratic programming.
A semidefinite programming problem is a convex constrained optimization problem where one optimizes a symmetric n × n matrix of variables X:
<> 代表内积。
SDP is a generalised form of linear program on X with infinite many linear constrains: