this week in SimPEG
week of June 12
This week in SimPEG…
Rowan Cockett was quite productive! 2 updates from him
A couple of his favourite lines are
If we used a finite difference algorithm to explicitly calculate each of the 112,500 columns of the Jacobian with a simple forward difference, we would require a calculation for each model parameter — or approximately 8.5 years of computational time. Furthermore, we would need to recompute the Jacobian at each iteration of the optimization algorithm. In contrast, if we use the implicit sensitivity algorithm presented in this paper, we can solve the entire inverse problem in 34.5 hours.
For this mesh, inverting for five spatially distributed parameters requires over 100 times less memory when using the implicit sensitivity algorithm, allowing these calculations to be run on modest computational resources.
(2) Rowan has a complete draft of his thesis. Bam!
Dom Fournier has been making progress on including rotated gradients in SimPEG: “Week 24 — and we’ve reached symmetry…”
As for an update from me: with SEOGI KANG, Rowan Cockett and Doug Oldenburg, we have been working on writing a tutorial on the 1D Magnetotelluric inverse problem. We are polishing off a set of notebooks that accompany the tutorials. If you would like a preview, they are here: https://github.com/simpeg/tle-magnetotelluric_inversion
In the meeting this week, we chatted through a few issues with the usability of the regularization implementation and chatted through some updates coming in through em/dev.
See you next week!