search

SciML / DiffEqFlux.jl

Pre-built implicit layer architectures with O(1) backprop, GPUs, and stiff+non-stiff DE solvers, demonstrating scientific machine learning (SciML) and physics-informed machine learning methods
https://docs.sciml.ai/DiffEqFlux/stable
MIT License
858 stars 154 forks source link

Suggestion: Apply to infection prediction #607

Closed EngrStudent closed 9 months ago

EngrStudent commented 3 years ago

My kids currently (~ww33'21) have hand-foot-mouth and I was looking at the papers and articles on the subject to try and get vague sense of how long they will be infectious.

In the exploration I saw this: https://www.nature.com/articles/s41598-020-68840-3

It is in Nature, a relatively credible journal, and from middle of last year, so relatively relatively recent.

The best they had was a deep recurrent convolutional residual (homebrew) something.
And the basic SIR models are differential equations, and they could do a job here.

I do not understand the neural+ODE and the SEIR epidemiology well enough to do a decent job, but I strongly suspect that a non-shallow Neural ODE based on SEIR models might substantially outperform the homebrew substantially. If they could compare favorably against relatively recent, relatively credible approaches that are being shared, then I think they should be considered.

If credible neural+ODE approaches for SEIR models have been articulated, could you point me to them. If not, then you might consider making the approach known to the corresponding author, and considering engaging in follow-up there.

Regards, Michael Munroe (engineering student)

ChrisRackauckas commented 3 years ago

Yes, this could be an interesting project for someone to pick up.

ChrisRackauckas commented 9 months ago

https://www.nature.com/articles/s41598-023-37118-9 https://www.cell.com/patterns/pdf/S2666-3899(21)00034-9.pdf https://www.osti.gov/biblio/1888443 https://spj.science.org/doi/full/10.34133/2021/9798302?adobe_mc=MCMID%3D22000014905405683995335849378418609464%7CMCORGID%3D242B6472541199F70A4C98A6%2540AdobeOrg%7CTS%3D1674864000 https://www.cell.com/patterns/pdf/S2666-3899(20)30193-8.pdf

EngrStudent commented 9 months ago

Very nice! Thank you.