doc_custom_nn.md

Specify neural network in command line

To use a custom neural network in command line with DomainLab, the user has to implement the following signature in a python file and specify the file path via --npath

def build_feat_extract_net(dim_y, remove_last_layer=False):

The user could choose to ignore argument remove_last_layer since this argument is only used in fair benchmark comparison.

See examples below from --npath=examples/nets/resnet.py where the examples can be found in the examples folder of the code repository. https://github.com/marrlab/DomainLab/blob/master/examples/nets/resnet.py

Example use case

model 'erm' with custom neural network

python main_out.py --te_d=caltech --task=mini_vlcs --debug --bs=2 --model=erm --npath=examples/nets/resnet.py

trainer 'matchdg' with custom neural network

python main_out.py --te_d=caltech --task=mini_vlcs --debug --bs=2 --model=erm --trainer=matchdg --epochs_ctr=3 --epos=6 --npath=examples/nets/resnet.py

model erm with custom neural network

python main_out.py --te_d=caltech --task=mini_vlcs --debug --bs=8 --model=erm --npath=examples/nets/resnet.py

Larger images:

model erm with implemented neural network

python main_out.py --te_d=caltech --task=mini_vlcs --debug --bs=8 --model=erm --nname=alexnet

model dann with implemented neural network

python main_out.py --te_d=caltech --task=mini_vlcs --debug --bs=8 --model=dann --nname=alexnet

Custom algorithm defined in external python file

python main_out.py --te_d=caltech --task=mini_vlcs --debug --bs=3 --apath=examples/models/demo_custom_model.py --model=custom --nname_argna2val my_custom_arg_name --nname_argna2val alexnet

python main_out.py --te_d=caltech --task=mini_vlcs --debug --bs=3 --apath=examples/models/demo_custom_model.py --model=custom --npath_argna2val my_custom_arg_name --npath_argna2val examples/nets/resnet.py