I always come back to @romainbrette.bsky.social's point about the difference between details vs realism in models journals.plos.org/ploscompbiol... - ThreadSky

About ThreadSky

cianodonnell.bsky.social • 47 days ago

I always come back to @romainbrette.bsky.social's point about the difference between details vs realism in models

https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004114

Comments

bellecguill.bsky.social•47 days ago

It also inspired me, I think we go further. Can the "plane model" predict a motor disruption? wind perturbation?

The wooden model is useless. But the paper plane needs an upgrade to make a prediction.

An ML model needs a mechanistic upgrade (or data, it's another topic) to predict perturbations

cianodonnell.bsky.social•47 days ago

Do you think you could guess which mechanisms are needed a priori?

bellecguill.bsky.social•47 days ago

It's hard, but I think it's what many modelers are trying to do. But the conclusion is sometimes subjective or speculative, until it accumulates evidence for it.

I think "perturbations testing" emerges as a good way to validate mechanisms inside data-constrained models.

cianodonnell.bsky.social•46 days ago

Agree it's a good test. Would you say perturbations are more special than other types of 'unseen data' though?

dylanmuir.bsky.social•46 days ago

IMO if a model can replicate a perturbation, this says something about a causal mechanistic similarity between three model and the real system. “Unseen data” is also a good test, but is less tied to mechanism.

cianodonnell.bsky.social•46 days ago

Sounds plausible... but why? Both are generalisations

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply