查看“︁相关向量机”︁的源代码

{{机器学习导航栏}}

'''相关向量机'''（Relevance vector machine，RVM）是使用[[贝叶斯推理]]得到[[回归]]和[[分类]]的[[简约]]解的[[机器学习]]技术。RVM的函数形式与[[支持向量机]]相同，但是可以提供概率分类。

其与带[[协方差函数]]的[[高斯过程]]等效。：
:<math>k(\mathbf{x},\mathbf{x'}) = \sum_{j=1}^N \frac{1}{\alpha_j} \phi(\mathbf{x},\mathbf{x}_j)\phi(\mathbf{x}',\mathbf{x}_j) </math>
其中φ是[[核函数]]（通常是高斯核函数），'''x'''<sub>1</sub>,…,'''x'''<sub>''N''</sub>是[[训练集]]的输入向量。{{Citation needed|date=February 2010}}

Compared to the SVM the Bayesian formulation allows avoiding the set of free parameters that the SVM has and that usually require cross-validation based post optimizations. However RVMs use an [[Expectation Maximization]] (EM)-like learning method and are therefore at risk of local minima, unlike the standard [[Sequential Minimal Optimization|SMO]]-based algorithms employed by [[Support vector machine|SVM]]s which are guaranteed to find a global optimum.{{Citation needed|date=February 2010}}

== 参考 ==
* {{cite journal |last=Tipping |first=Michael E. |title=Sparse Bayesian Learning and the Relevance Vector Machine |year=2001 |journal=[[Journal of Machine Learning Research]] |volume=1 |pages=211–244 |url=http://jmlr.csail.mit.edu/papers/v1/tipping01a.html |doi=10.1162/15324430152748236 |access-date=2010-03-31 |archive-date=2020-02-19 |archive-url=https://web.archive.org/web/20200219044718/http://jmlr.csail.mit.edu/papers/v1/tipping01a.html |dead-url=no }}

== 软件==
* [http://dlib.net dlib C++ Library] {{Wayback|url=http://dlib.net/ |date=20200917123233 }}
* [http://www.terborg.net/research/kml/ The Kernel-Machine Library] {{Wayback|url=http://www.terborg.net/research/kml/ |date=20200813210406 }} 

==外部链接==
*[http://www.relevancevector.com Tipping's webpage on Sparse Bayesian Models and the RVM] {{Wayback|url=http://www.relevancevector.com/ |date=20080704175611 }}
*[https://web.archive.org/web/20111005202038/http://www.tristanfletcher.co.uk/RVM%20Explained.pdf A Tutorial on RVM by Tristan Fletcher]

[[Category:分類演算法]]
[[Category:Ensemble learning]]
[[Category:机器学习]]
[[Category:Non-parametric Bayesian methods]]