SimLearn

Completing Training Data by Iteratively Learning Simulation

Machine learning methods based on existing training data have proven to be very effective in identifying patterns and implicit dependencies in complex situations with many parameters and in providing classification, prediction and decision support with the models learned. In practice, however, the large amounts of correctly labeled training data required for such approaches are often not available.

Based on actual application examples from the agricultural sector, SimLearn examines the suitability of a new approach in which existing operative knowledge codified in simulation models is combined iteratively with the increasing insights of learned models: Extensive synthetic training data sets are generated by existing simulation models. A learning system initiated on such data will then be extended and improved by empirical data collected of actual farms. This combination fills gaps in the existing database and enables improved training. The result is a learned, more powerful model of the observed reality with improved usage potentials.

SimLearn exemplary considers the operational decisions in crop production on operational and tactical level with regard to income and environmental effects. The bioeconomic modeling system MPMAS_XN of Hohenheim University (UHOH) allows initial simulations of the effects of fertilization and cultivation decisions both from a biological (plant growth) and an economic (expected revenue) point of view. This information is combined and compared with the results of cooperating experimental farms and with standard and average values from the databases of the Kuratorium für Technologie und Bauwesen in der Landwirtschaft (KTBL). Using those generated data collections, DFKI iteratively trains a suitable learning system which enables an improved prediction and assessment of alternative courses of action.

Period: 01.01.2020 - 30.06.2023
Funding: 1,87 Mio EUR (total), 0,51 Mio EUR (UHOH), BMBF / DLR