A dummy's guide to Physics-Informed Neural Networks (PINNs).

There is really not that much to unpack in PINNs. Their popularity is a travesty of computational science and engineering.
Watch out for a short video series with tutorials and paper reviews.

My views on Scientific Foundation models.

We are hiring a postdoc researcher! Please share!

As the "Research Fellow in Computational Fluid-Structure Interaction", you will work on the MAPFSI project funded by the EPSRC, developing cutting-edge computational algorithms for challenging FSI problems.

1/

#AI for sustainability is like
#pollution for clean air
🤣🤣

Please share!

We are now accepting applications for the fully-funded PhD position (applicable worldwide) on Computational Methods for Fluid-Structure Interaction problems.

The details of the PhD studentship are available at
findaphd.com/phds/project...

1/3

They went from "data-driven" to "data-free" in a couple of years,
says data-free, but needs the desired solution at a finite number of points...
uses FEM as a replacement for automatic differentiation 🤣

they are circling back 😂

#SciML is crazily bonkers!

A bad way to start a course on #HPC is to throw a bunch of slides with hardware & specifications at the audience and kill their enthusiasm for using the HPC.

Start with hands-on activities like accessing HPC and running some simple programs first. Then, show the hardware later.

Just leaving this here 😭😭

This is beyond ridiculous!

Happy New Year, everyone! 🎉🍾

Learn more.
Publish less.
Help/guide/mentor others.

Special thanks to all those who helped me and supported me. 🙏🙏

#Happy2025

I have documented my experience with PINNs.
Basically, PINNs are a highly convoluted way of solving PDEs using collocation least-squares, which we abandoned because of their poor performance.

I hope you find this useful. Let me know what you think.
Link: researchgate.net/publication/38…