Research Article

Linearized and Kernelized Sparse Multitask Learning for Predicting Cognitive Outcomes in Alzheimer’s Disease

Table 2

Performance comparison of various methods in terms of CC and wR on 10 cross validation cognitive prediction tasks.

Method ADAS MMSE RAVLT
TOTAL TOT6 T30 RECOG

Ridge0.603 0.0310.407 0.0400.401 0.0840.361 0.0920.377 0.0960.261 0.080
Lasso0.655 0.0360.540 0.0460.493 0.0840.507 0.1000.523 0.1060.416 0.087
MKL0.658 0.0300.544 0.0520.502 0.0660.476 0.0950.506 0.1050.391 0.072
Robust MTL0.587 0.0220.338 0.0840.423 0.0900.432 0.0960.444 0.0940.354 0.105
CMTL0.603 0.0250.381 0.0420.397 0.0720.362 0.0900.381 0.0990.260 0.068
Trace0.548 0.0390.144 0.0910.342 0.1720.395 0.1590.402 0.1420.253 0.130
SRMTL0.655 0.034 0.525 0.0580.492 0.0790.505 0.0970.523 0.1030.413 0.092
-MTL0.662 0.0430.532 0.0560.532 0.0820.492 0.1090.522 0.1050.404 0.091
-MKMTL0.661 0.0340.460 0.0990.519 0.0720.470 0.0890.494 0.0940.412 0.090
-MKMTL0.660 0.0350.547 0.0450.529 0.0790.500 0.0950.508 0.0940.421 0.075

MethodFLUTRAILSwR
ANIMVEGAB

Ridge0.185 0.0900.396 0.0730.291 0.0970.330 0.1100.361 0.041
Lasso0.365 0.0960.506 0.0590.363 0.0410.467 0.0960.484 0.049
MKL0.375 0.0710.496 0.0670.374 0.0560.457 0.0600.478 0.046
Robust MTL0.253 0.0960.443 0.0570.282 0.1130.292 0.1230.385 0.038
CMTL0.180 0.0890.390 0.0710.287 0.1160.335 0.1120.358 0.036
Trace0.212 0.1430.331 0.1120.270 0.1120.290 0.1220.319 0.083
SRMTL0.362 0.0930.503 0.0640.340 0.0630.361 0.0950.468 0.045
-MTL0.379 0.0760.501 0.0630.399 0.0600.467 0.0980.489 0.050
-MKMTL0.381 0.0800.521 0.0670.421 0.0640.481 0.0760.482 0.047
-MKMTL0.409 0.0730.516 0.0650.417 0.0670.490 0.0870.500 ± 0.043