Thursday, September 21, 2017 at 12:20pm to 1:10pm
Humans learn from experience, whereas machines need to be told what to do. Can we get machines to learn from experience? The answer is yes! That is precisely what machine learning is. Machine experience has a name and is called “data”. Now it is possible for machines to solve problems - when the machine has been trained with large datasets and with advanced statistical predictive models. Self-driving cars, spam email identification, practical speech recognition, effective web searches, and understanding of the human genome are perfect examples of applied machine learning. Recently, the growth of digital agriculture and its related technologies has created a vast amount of new data opportunities. Application of learner-based predictive models has created promising opportunities in digital agricultural research mostly in remote and proximal sensing, yield prediction, fertilizer recommendation and digital soil mapping. This seminar will present findings from two applications of machine learning in digital agriculture: (1) spatial prediction of maize fertilizer requirements in Bangladesh using Random Forest Regression, and (2) meta-ensemble machine learning algorithms for spatial interpolation of soil organic carbon in the western USA.