您好 有一个问题您能给解答一下吗 about udtl

[jy2020-start]

okay Thank you, your answer give me a lot of help

[chandu07]

@jy2020-start I used google translate and if it translated the question right, this is what you can do. You can provide the --transfer_task argument with the list of domains you want to consider for training and testing. For example, in case of PU dataset, if you want it to consider 0 and 1 domains for training and 2 and 3 for testing, input --transfer_task as [[0,1],[2,3]].

python train_advanced.py --transfer_task [[0,1],[2,3]] will run the default model with the multi domain training and adaptation for the above explained example.

[jy2020-start]

Thank you for your explanation，I understand your explanation of MULTIDOMAIN UDTL, but what are the command code parameters in case of LABEL-INCONSISTENT UDTL。

[chandu07]

I missed that part, Label-Inconsistent version of UDTL isn't implemented here. It is linked separately and is also mentioned in this repo's description. You can refer to this.