a) For classification tasks, the data needs to be split in training and test sets to optimize the param- eters. True; otherwise over-fitting might occourr. b) The test error is always higher than the training error. False, but it is very likely that the test-error is higher. c) The more genes I use for training a classifier, the better will be its performance. False. When using higher dimensional datasets clustering artefacts will occour, which lower the performance.