search

InstituteforDiseaseModeling / laser

Light Agent Spatial modeling for ERadication
MIT License
2 stars 5 forks source link

Normalize gravity/network matrix by row rather than globally #37

Open clorton opened 3 weeks ago

clorton commented 3 weeks ago

Normalizing the network "globally" effective zeros out or cuts off the "tail" of the network connections. See @krosenfeld-IDM 's work in the GEOMED repository.

KevinMcCarthyAtIDM commented 3 weeks ago

In the migration models Jupyter notebook I put together and shared, there should be utility functions to use a number of different normalization or limiting options - rescaling the whole matrix to set "average outbound migration" to some number (to make the "k" parameter more interpretable); rescaling row-by-row to limit maximum outbound migration from highly connected nodes but leaving rows below that maximum alone; and also rescaling inbound migration to limit inbound migration to a node to some fraction of the node's population (not sure when this would ever be useful, though).