On-Demand ML model development
The need was development of a route creater which can adapt new scenerios which is not defined or unknown.
Due to unknown nature of problem, model was in need of understanding changing situations and adapt to them. There were not any scenarios defined, model was exploring them by itself and after a change in scenario adapt to new scenario. Below you see adaptation graphs. Since this is core AI development there is not any application developed.
Below you see the accuracy of the models. Bottom is time and left is accuracy. Colors of the lines is the hyper-parameters of the model. These 2 graph are belong to different model approaches, and they require different hardware settings. The idea is using appropriate model with true hyper-parameter in different cases. Here you see accuracy is around %90 after the learning operation completed and it goes stable afterwards.
