Algorithms
The overarching goal of this study is to introduce a framework and method for inferring model parameters by integrating non-traditional model drivers and surrogates as inputs during the model calibration process and intervention planning.
The Optimization Engine runs the different algorithms added and exposed in the accelerated decision making with AI tool. In calibration, the environment gives the action space and output/observation space to the algorithm which performs optimization based on the inputs.
Multiple algorithms can be implemented and onboarded in the tool and coupled with different models and environments:
At present we have implemented:
- Reinforcement Learning- Bayesian Optimization- Genetic Algorithms- Simplex Methods
To perform either model calibration or intervention planning with these algorithms, the models must be wrapped in an environment.