Report of Experiences

• All things considered, progress was good, and goals were met.

• Needed to choose a Gaussian white noise generation method.

  • Central limit theorem approach vs. rejection sampling

  • Central limit theorem

    • Need to sum 50 random numbers for each sample

  • Rejection sampling

    • Need to cast 10 pairs of random numbers for each sample

    • For each pair, need to evaluate the Gaussian 1D function

    • Slower than Central limit theorem approach

• Many small problems with CSIL machines, which hang or reboot without warning.

• Main code runs on CSIL, but renderings & plots can only be done in the GrUVi lab. So, ssh'ing was required and it was not possible to fully automate tasks.