Comments (2)
  • Add a Comment
  • Edit
  • More Actions v
  • Quarantine this Entry

1 PaulRubin commented Permalink

For a model whose parameters are estimated/forecast/snatched from a place the moon don't shine, obsessing about numerical accuracy (or, with a MIP model, squeezing the gap down to zero) is probably not energy well spent.

2 JeanFrancoisPuget commented Permalink

Right, this was Samik's point I guess. My point was to not lose track of cases where accuracy is required.