So I’m working on a project that has 3 datasets. A dataset connectome data extracted from MRIs, a continuous values dataset for patient scores and a qualitative patient survey dataset.<\/p>\n
The output is multioutput. One output is ADHD diagnosis and the other is patient sex(male or female).<\/p>\n
I’m trying to use a gcn(or maybe even other types of gnn) for the connectome data which is basically a graph. I’m thinking about training a gnn on the connectome data with only 1 of the 2 outputs and get embeddings to merge with the other 2 datasets using something like an mlp.<\/p>\n
Any other ways I could explore?<\/p>\n
Also do you know what other models I could you on this type of data? If you’re interested the dataset is from a kaggle competition called WIDS datathon. I’m also using optuna for hyper parameters optimization.<\/p>\n<\/div>\n
submitted by \/u\/Luccy_33 <\/a>
[link]<\/a><\/span> [comments]<\/a><\/span><\/p>
