I found a few mistakes in Pseudocodes/VFI_PseudoCodes.pdf at main · vfitoolkit/Pseudocodes · GitHub, Section 2.1.1 Refinement.
Preliminary step. We need also a loop over a^\prime since we maximize over d for all (a',a,z)
The policy for a^\prime is sometimes denoted g^{a^\prime}_n, other times g_n
It would be useful to let an AI scan the pseudocode document for inconsistent notation and imprecisions.
Hopefully fixed them all
Did you reconpile the tex? The two points I flagged in my post are not fixed yet.
The AI wasn’t as smart as you so it missed that one.
Fixed now
Related to the pseudo-code on grid interpolation: I noticed that the explanation of linear interpolation for finite horizon VFI is no longer present in GitHub - vfitoolkit/Pseudocodes: Pseudocodes for the commands of VFI Toolkit · GitHub
I found a really nice explanation of how the toolkit uses linear interpolation in VFI in GitHub - robertdkirkby/GridInterpolationLayer: description of grid interpolation layer algo · GitHub. The figures are great!
The pseudo-code for grid interpolation with finite-horizon vfi is good but the section with decision variable is missing. It was instead present in a previous version of GitHub - vfitoolkit/Pseudocodes: Pseudocodes for the commands of VFI Toolkit · GitHub
To sum up, I have a small request: is it possible to show somewhere the pseudo-code for “finite-horizon with d variable” plus “gridinterplayer”? This helps users (and AI tools as well) to better understand the code.
I put links to grid interpolation layer and divide and conquer pdfs/repos in the pseudo-codes pdf.
Added the ‘conditional on d’ explanation to the grid interpolation layer pdf.
