Fig. 1
From: Interpretable and integrative deep learning for discovering brain-behaviour associations
Interpretation framework based on the DAA. The three parts of the process are represented by the three columns. The DAA column shows candidate associations (links between brain and behaviour) generated by the DAAs. A hierarchical bootstraping procedure organises the following parts. The r-DAA column shows the ensembling step performed on associations derived from model communities (community size ranging from 1 to \(n_E\)) trained with the same training data but with different initialisations. Finally, the stability selection column shows the construction of stability paths parameterised by \(n_E\). From the generated N binary decision support, a stability selection procedure is used to retain associations with stability path \(\Pi \ge \pi _\text {thr}\).
