This thread is for anyone working on personal projects to share their progress, and hold themselves somewhat accountable to a group of peers.
Post your project, your progress from last week, and what you hope to accomplish this week.
If you want to be pinged with a reminder asking about your project, let me know, and I'll harass you each week until you cancel the service.
Jump in the discussion.
No email address required.
Notes -
A coworker posted this competition RealPDE, and I signed up. Though I currently snoozed past the first few weeks.
It's funny that the Navier-Stokes 3D equation proof just got solved via an LLM, its only tangentially related but funny because I was just considering it as a potential PDE. I built out my training pipeline this weekend and did a marble test, but I still need to design my model/loss approach. I was originally thinking of doing a physics-informed-neural-network(PINN) but after looking through it my understanding is that they give us 3D airfoil cross sectionals meaning the actual data follows 3D flow but we only have the 2D slice from PIV. The 2D and 3D Navier-Stokes are sufficiently different from each other that using the 2D as a loss function a-la PINNs is not going to generalize to a 3D real system, and the 3D NS PDE is going to require information that the training/validation data lacks. Meaning a model probably needs to learn everything that the 2D slice doesn't capture(?).
So now I need to read up on physics informed neural operators (PINOs). I really want to explore the Causal Transformer direction, maybe combine it with a PINO since I'm not a mechanical engineer or mathematician and the idea of a SCM applied to DL sounds like it would have interesting applications to me.
I'm unsure if this is sufficiently in-scope for this thread, but considering that I'm probably going to spend a significant amount of free time on this for the next month, I'd classify it as a hobby.
More options
Context Copy link