Ideas for a non intrusive gradient descent optimizer

   347   0   1
User Avatar
Member
35 posts
Joined: Dec. 2014
Offline
Say you have some nodes in a subnet, and a target geometry

Without doing any modifications to the subnet, i.e non intrusively, how can you get a set of values for each parameter of each node in the subnet, or of only a subset of the parameters / nodes, that minimizes the distance to the target.

Does it sound like a good task for APEX, Attributes from Parameters, Invoke Compiled Graph, and a bit of metaprogramming to handle the computation graph from AFP ?
Edited by alecaille - July 17, 2026 06:27:14
  • Quick Links