In the trcrpm paper the temporal cohesion function G is used to modify the CRP probability assignment for new state assignments z_t. Appendix A to the paper defines G as a student T reweighting function.
We want to isolate for the behaviour of G, but when we step through the trcrpm code - we are finding that G is less explicitly coded and more a convenient by product of CGPM/Crosscat and the "views" that are setup on the incoming data.
We would like to modify G to another form... we are wondering if you could suggest a convenient location in the code for this.
In the trcrpm paper the temporal cohesion function G is used to modify the CRP probability assignment for new state assignments z_t. Appendix A to the paper defines G as a student T reweighting function.
We want to isolate for the behaviour of G, but when we step through the trcrpm code - we are finding that G is less explicitly coded and more a convenient by product of CGPM/Crosscat and the "views" that are setup on the incoming data.
We would like to modify G to another form... we are wondering if you could suggest a convenient location in the code for this.