Behnam Kazemivash

A single learned model of brain dynamics yields both a time-averaged connectivity representation and a trajectory-geometry representation, and which one carries condition-discriminative information shifts with chronnectome complexity. On eight conditions in the Human Connectome Project young adult cohort, the full model reached 97.4% classification accuracy; with the time-averaged representation zeroed out, trajectory-only accuracy ranged from 93.4% for Working Memory to 0.0% for Relational Reasoning, the latter indistinguishable from a label-permutation null. In the aging cohort's four conditions, the time-averaged representation carried most of the discriminative information, yet the contrast between higher- and lower-complexity conditions persisted.

A single learned model of brain dynamics yields both a time-averaged connectivity representation and a trajectory-geometry representation, and which one carries condition-discriminative information shifts with chronnectome complexity. On eight conditions in the Human Connectome Project young adult cohort, the full model reached 97.4% classification accuracy; with the time-averaged representation zeroed out, trajectory-only accuracy ranged from 93.4% for Working Memory to 0.0% for Relational Reasoning, the latter indistinguishable from a label-permutation null. In the aging cohort's four conditions, the time-averaged representation carried most of the discriminative information, yet the contrast between higher- and lower-complexity conditions persisted.