luiarthur / CytofRepFAM.jl

Research code for repulsive feature allocation model for mass cytometry data.
MIT License
0 stars 0 forks source link

Simulation Study 6.8.1 and Real Data Analysis #8

Closed luiarthur closed 4 years ago

luiarthur commented 4 years ago
  1. use iMCMC, K=10, L=3
  2. use parallel tempering with temperatures {1, 1.003, 1.006, 1.01}
  3. True Z = {Z1, Z2, Z3}
  4. With missing data, and without missing data
  5. Phi=(0, 1, 10, 25)

Points 3-4 imply that there are 6 datasets. Point 5 suggests we need to analyze those 6 datasets 4 times (with different phi). That's 24 runs. Each run using 4 cores. If I use 24 cores on the server, that means I can do 6 runs (with PT using 4 cores) at a time. And I would need to iterate 4 times. Probably looping over phi. If one run takes 15 hours, then *total run time = 15 4 = 60 hours**.

That's 4 runs, each using 4 cores. This requires 16 cores. So, this can be run in parallel.