Skip to main content
. 2020 Jun 2;9:e56601. doi: 10.7554/eLife.56601

Figure 3. Searchlight hyperalignment for building a common model and transformation matrices for the whole cortex.

Figure 3.

Overlapping transformation matrices (Rij, Rik, …) are calculated for overlapping searchlights in each subject and aggregated into a single whole-cortex transformation matrix, RiA. The transpose of the transformation matrix, RiAT, allows pattern vectors from the common model space to be projected into an individual brain’s topographies. The aggregation step results in a whole-cortex transformation that is mostly orthogonal locally (Guntupalli et al., 2016) but globally nonorthogonal.