Trees to Flows and Back: Unifying Decision Trees and Diffusion Models(arxiv.org)

54 pointsby rsn243الشهر الماضي11 comments

Trees to Flows and Back: Unifying Decision Trees and Diffusion Models

https://arxiv.org/abs/2605.00414

11 comments

niksmatherالشهر الماضي

Apologies if I didn't understand the paper, but why do you want to apply diffusion models to tabular datasets in the first place?

Do we think they'll be better than decision trees? Is there some tabular problem that can be handled by diffusion but not trees?

henrydarkالشهر الماضي

First, they give a novel generation algorithm based on combining trees with diffusion, which trees alone just don't give you.

Second, yes, they think some tabular data will be fit better by their combination of trees with diffusion than just with trees.

robotresearcherالشهر الماضي

You might not want to make a sword out of iron if steel is available, but understanding the relationship between iron and steel is broadly valuable.

niksmatherالشهر الماضي

I can see the mathematical results are interesting, I was more wondering if there was a practical utility to this TreeFlow thing they built.

semessierالشهر الماضي

this lacks the math for any bold claims

emil-lpالشهر الماضي

Did you read the paper? Is there something specifically you're missing? A proof? A theorem statement?

semessierالشهر الماضي

this is an empirical engineering paper with theoretical dressing, it would not need to be a theorem paper of course.

henrydarkالشهر الماضي

Is the code available somewhere?

rsn243الشهر الماضي

Decision trees and diffusion models are ostensibly disparate model classes, one discrete and hierarchical, the other continuous and dynamic. This work unifies the two by establishing a crisp mathematical correspondence between hierarchical decision trees and diffusion processes in appropriate limiting regimes. Our unification reveals a shared optimization principle: \emph{Global Trajectory Score Matching (GTSM)}, for which gradient boosting (in an idealized version) is asymptotically optimal. We underscore the conceptual value of our work through two key practical instantiations: \treeflow, which achieves competitive generation quality on tabular data with higher fidelity and a 2\times computational speedup, and \dsmtree, a novel distillation method that transfers hierarchical decision logic into neural networks, matching teacher performance within 2\% on many benchmarks.

Jaxanالشهر الماضي

You could at least fix the latex commands when copy pasting the abstract. ;-)

goroldالشهر الماضي

Figure 1 definitely cleared up any misunderstandings I had about the paper