This guide provides instructions on accessing and using the AV-MNIST dataset and the MIMIC-III dataset for multimodal learning.
Download the avmnist.tar.gz
file from this link.
The available command-line options for AV-MNIST include:
--fusion_type=
: Choose the fusion method with optionslf
(late fusion) orlrtf
(low-rank tensor fusion).--model_type=
: Select the modeling approach fromunimodal
,inter_modality
,intra_modality
, andinter_and_intra_modality
.--test
: Evaluate the model performance.
python avmnist_main.py --model_type="inter_and_intra_modality" --fusion_type='lf'
Access to MIMIC-III requires authorization. Follow the instructions here to obtain necessary credentials.
The command-line options available for MIMIC-III include:
--task=
: Specifies the task to be performed. The available options aremortality
,mimic_1
,mimic_7
.--model_type=
: Defines the model's type. Options includeunimodal
,inter_modality
,intra_modality
,inter_and_intra_modality
.--test
: Activates evaluation mode for the trained model.
python your_program.py --task="mortality" --model_type="inter_and_intra_modality"
The code is build upon pliang279/MultiBench.
We thank the authors for their amazing work and releasing the code base.