| More powerful "Automatic Experiments" now available [message #1830504] |
Mon, 27 July 2020 23:47 |
 James Kaufman
Messages: 240
Registered: July 2009 |
Senior Member |
|
|
Automatic Experiments use a Nelder-Meade downhill simplex algorithms to automatically optimize any disease model with respect to reference data. Previously, AI used an error function based only on reference daily incidence data. This is reported by public health. In response to the pandemic, many users would like to optimize their models with respect to cumulative incidence, daily incidence, and or cumulative deaths (or some combination of the three).
Now available in the latest integration build a new compound error function has been added. Users can optionally select this error function. It provides capability to select and fit, any of the following combinations:
1. Daily Incidence
2. Cumulative Incidence
3. Cumulative Deaths.
Ablation testing was done using this new error function on hand-selected regions where we were observing issues using only the simple error function. The efficacy of this new implementation was documented and found to be correct. Please see the new doc at:
https://wiki.eclipse.org/Running_an_Automated_Experiment
|
|
|
Powered by
FUDForum. Page generated in 0.02840 seconds
] 
Search
Help
Register
Login
Home
