Research Article

Efficient Multiple Kernel Learning Algorithms Using Low-Rank Representation

Table 1

The performances of MKL algorithms and LR-MKL algorithms on the datasets Yale, ORL, LSVT, and Digit.

ā€‰YaleORLLSVTDigit
ā€‰AccTimeAccTimeAccTimeAccTime

SVM(best)69.32310.298180.00021.679876.19050.014295.93020.9195
LR-SVM(best)79.86670.010987.00150.157577.66370.004396.33040.0972
UMKL(+) [16]87.51142.598193.579818.235278.57140.022997.35714.7412
LR-UMKL(+)95.89360.301883.48131.081773.80950.015298.15952.0278
UMKL() [16]58.32482.724477.502220.550766.69350.028196.09047.1661
LR-UMKL()93.57390.263693.40632.183667.00330.017698.46183.6392
AMKL [17]85.77533.723693.87414.685480.95240.045297.472511.2138
LR-AMKL94.37990.386996.94170.459288.09520.008598.69526.8804
GMKL [18]86.29894.533096.20575.007085.71430.056599.14998.0774
LR-GMKL95.80150.625398.58330.776188.92860.018399.46734.5972
LMKL(sof) [19]87.9077215.405597.0003220.312285.00905.198999.8898166.7978
LR-LMKL(sof)97.935222.720498.972417.379186.74291.193398.359198.1812
LMKL(sig) [19]88.0145106.755297.0108107.091188.75410.723899.375048.5914
LR-LMKL(sig)98.066715.971197.997911.724092.66270.459499.562524.5927
HMKL [20]63.603792.441093.5109118.234080.59980.091597.625810.3559
LR-HMKL91.96118.597298.689310.936885.16250.035298.34796.3959
CMKL [21]86.416695.061896.0308107.694079.95030.087496.501410.6074
LR-CMKL94.008310.638098.479912.974693.90240.034798.91136.2696
PMKL() [22]89.00356.184298.49016.906592.85710.107999.588124.8702
LR-PMKL()99.14290.905399.57120.982895.93860.0573100.000012.9831
PMKL() [22]89.02615.389398.75336.545092.46620.129599.504621.1679
LR-PMKL()98.89680.849499.58280.891195.51450.065199.794113.5108
ANMKL() [23, 24]86.72106.485698.439620.456491.98270.116798.000710.3979
LR-ANMKL()96.86670.704199.46432.851992.22470.024799.28506.9070
ANMKL() [23, 24]86.69987.066498.220421.161593.00350.119498.00399.7753
LR-ANMKL()97.29170.886399.28573.059792.53910.022499.24975.0374