liunelson
commented
1 year ago Risk-Based Optimization Under Uncertainty (RBOUU), a.k.a. "optimize"
This is a feature in PyCIEMSS that takes a model, a quantity of interest, and some interventions (i.e. setting some variables/parameters to some to-be-determined values at some given times) and returns an "optimal policy" (i.e. values of the interventions such that (1) an objective function over the intervention variables/parameters and (2) the quantity of interest are simultaneously minimized.
Example: given a SIDARTHE model, the 2-last-days average of I as the quantity of interest, and an intervention of beta \in [0.0, 3.0] at time t = 0.1, we get the optimal policy of interventionbeta = 0.0.
Test notebook: https://github.com/liunelson/pyciemss/blob/nl-test/notebook/integration_demo/nliu/review_interfaces.py
@fivegrant here's a handy table that I made, where I describe which/how each argument should be exposed for integration. https://docs.google.com/spreadsheets/d/1_iyt9TFcdUlTzYOebhk2_Kz-ToW6wJFXQ3Vinbovwx8
@jamiewaese-uncharted @pascaleproulx ^ this table could serve as guide to the design of the workflow boxes.
Work on optimization techniques considering uncertainties in the system.