User Tools

Site Tools


start

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
start [2016/02/25 10:41]
josh Added note about averaging of independent runs.
start [2016/03/03 13:13] (current)
josh
Line 28: Line 28:
 FRESHS is an open source project: if you have an algorithm which you would like to contribute, the harness has been designed such that this should be relatively easy for you to do, instantly making your method available to users of all the implemented simulation packages. We hope that in time FRESHS will become the platform of choice for validation, comparison and distribution of rare event sampling methods. FRESHS is an open source project: if you have an algorithm which you would like to contribute, the harness has been designed such that this should be relatively easy for you to do, instantly making your method available to users of all the implemented simulation packages. We hope that in time FRESHS will become the platform of choice for validation, comparison and distribution of rare event sampling methods.
  
- 
-===== Averaging Multiple Runs ===== 
- 
-The FFS and FFS-flatPERM algorithms continue sampling at each interface until a fixed number of shots progress to the next interface. This type of trial produces successes according to a negative binomial distribution:​ for a single trial, the success probability is estimated simply as the proportion of successful shots; however for multiple trials (multiple independent FFS/​FFS-flatPERM calculations) the correct averaging to determine the success probability is N_success / (<​N_fails>​ + N_success). ​ The number of failures must be saved and averaged in order to then find the expected success rate. 
- 
-When it is desired therefore to combine multiple independent simulations,​ the structure of the interfaces must be the same place for each calculation,​ and the shot databases post-processed to get the correct averages. ​ Automatic interface placement is therefore contraindicated in this case.  Only the probability of success (the total statistical weight) needs to be corrected for this effect; the proportional statistical weights of individual trajectories relative to each other remain correct. 
  
  
start.txt ยท Last modified: 2016/03/03 13:13 by josh