An efficient radius-incorporated MKL algorithm for Alzheimer׳s disease prediction
作者：
Highlights：
• The objective of our L2BRMKL is convex and can be globally optimized.
• Our L2BRMKL achieves better classification accuracy by automatically tuning C.
• The objective of L2BRMKL is provably to be an upper bound of generalization error.
• Our L2BRMKL tends to select kernels with more discriminative power.
• Our L2BRMKL is more computationally efficient than other radius-incorporated ones.
摘要
Highlights•The objective of our L2BRMKL is convex and can be globally optimized.•Our L2BRMKL achieves better classification accuracy by automatically tuning C.•The objective of L2BRMKL is provably to be an upper bound of generalization error.•Our L2BRMKL tends to select kernels with more discriminative power.•Our L2BRMKL is more computationally efficient than other radius-incorporated ones.
论文关键词：Multiple kernel learning,Radius-margin bound,Support vector machines,Alzheimer׳s disease,Neuroimaging
论文评审过程：Received 28 January 2013, Revised 11 August 2014, Accepted 9 December 2014, Available online 17 December 2014.
论文官网地址：https://doi.org/10.1016/j.patcog.2014.12.007
