Interactive Machine Learning for Applications in Food Science and Technology
Résumé
The apparent simplicity of food processes often hides complex systems,where physical, chemical and living organisms’ processes co-exist and interact tocreate the final product. Data can be plagued byuncertainty;heterogeneityof avail-able information is likely;qualitativeandquantitativedata may also coexist in thesame process, from expert perception of food quality to nano-properties of ingredi-ents. In order to obtain reliable models, it then becomes necessary to acquire addi-tional information from external sources. Experts of a domain can provide invalu-able insight in products and processes, but this precious knowledge is often avail-able only in the form of intuition and implicit expertise. Including expert insightin a model can be tackled by having humans interacting with a machine learningprocess, through visualization or via specialists in encoding implicit domain knowl-edge. In this chapter, three selected case studies in food science portray differentsuccess stories of combining machine learning and expert interaction. We show thatexpert knowledge can be integrated at different stages of the modelling process cither online or offline, to initialize, enrich or guide this process.