Skip to content

2012 11 23

Jump to bottom Edit New page
kfl edited this page Nov 23, 2012 · 5 revisions

Agenda

  • Project direction: Vectorisation in Nikola - branch divergence avoidance is on hold (moved to "Future Work")
  • Experiments/benchmarks:
    • binomial
    • sobol
    • least squares
    • lsm
    • (gaussian)
  • Push-button plotting
  • Linear least squares regression
  • Our status: HMatrix linearSolveLS, Cholesky decomposition and QR decomposition doesn't seem appropriate
  • Rolf's R code? Where can we find it?
  • Root finding in Longstaff & Schwartz - we don't see the connection to least squares
  • Communication forms: Email, Trello & Github - when to use what.

Minutes

  • There is a strong connection between the experimental (e.g. vectorisation) work we select to do and which benchmarks we have selected. Our benchmarks should motivate the selection

    Ken: not motivate but support the selection.

    Next step: for each benchmark/case we should present how vectorisation will improve them.

  • We really shouldn't just trust the Python code for LSM - compare it with the binomial pricer. Also, Rolf has a LSM pricer on his website (see link in FAMØS article)

  • Least-Squares: Ken will give us pointers to some papers which contains linear least squares implementations.

  • We could improve our structure on Trello

Add a custom footer

Survey "VectorMARK" in progress

Clone this wiki locally