Closed AissamDjahnine closed 4 years ago
Very sorry for the late reply because the notification is blocked by the mailbox. For other datasets or tasks, you may refer to Issue #13 . Since there is a large domain gap between the natural image datasets and medical images, a carefully hyper-parameter tuning process is needed for a favorable result. Specifically, those parameters related to training network weights should be adjusted according to your dataset, and those related to architectural parameters learning can be less changed.
Hello, I would like to begin by thanking you for this work. Me as an intern , i am investigating P-DARTS ( and eventually PC-DARTS) for my breast-cancer diagnosis task and I am trying to find the best architecture possible for this.
And since all your tests were done on natural images datasets ( CIFAR AND IMAGENET ) ,do you have any comments about using them on such medical datasets? hyper-parameters that we should choose carefully ?
thanks in advance for answering.