the hyperparameters

[JINcan510]

Hello Author, thank you for your excellent article. I would like to ask if the best experimental result you obtained on ISIC-2017 in your paper (DSC 86.39) was achieved with the hyperparameters shown below? I look forward to your reply and would be extremely grateful!

parser = argparse.ArgumentParser()
parser.add_argument('--path_trained_models', default='./checkpoints/sup_xnet')
parser.add_argument('--path_seg_results', default='./seg_pred/sup_xnet')
parser.add_argument('--path_dataset', default='./dataset')
parser.add_argument('--dataset_name', default='ISIC-2017', help='CREMI, ISIC-2017, GlaS')
parser.add_argument('--input1', default='LL')
parser.add_argument('--input2', default='LH_HL_HH')
parser.add_argument('--sup_mark', default='100')
parser.add_argument('-b', '--batch_size', default=32, type=int)
parser.add_argument('-e', '--num_epochs', default=200, type=int)
parser.add_argument('-s', '--step_size', default=50, type=int)
parser.add_argument('-l', '--lr', default=0.05, type=float)
parser.add_argument('-g', '--gamma', default=0.1, type=float)
parser.add_argument('-u', '--unsup_weight', default=0.5, type=float)
parser.add_argument('--loss', default='dice', type=str)
parser.add_argument('-w', '--warm_up_duration', default=20)
parser.add_argument('--momentum', default=0.9, type=float)
parser.add_argument('--wd', default=-4, type=float, help='weight decay pow')

parser.add_argument('-i', '--display_iter', default=5, type=int)
parser.add_argument('-n', '--network', default='xnet', type=str)
parser.add_argument('--local_rank', default=-1, type=int)

[Yanfeng-Zhou]

I've graduated from my previous school and can't look up the specific hyperparameters I used, but from what I remember, -l, -g(05, 0.2, 0.1?), and -u(1, 3, 5?) are very important, and tuning them carefully will yield the results in the paper.

[JINcan510]

Okay, thank you very much for your reply. I wish you success in your career and happiness in life!