Model selection and optimization experiment of a field trial with agricultural data using Python

University essay from Lunds universitet/Matematik LTH

Abstract: Potato cultivation is vast for many agricultures in southern Sweden. What many might not be aware of, is how much pesticides which are used in the cultivation process. In "Late blight prediction and analysis"[1] predictions of late-blight attacks was modelled on potato-corps using Elastic Net among other methods. This thesis is an algorithmic complement describing Alternate Direction Method of Multipliers (ADMM) and its use for efficient optimization of Elastic Net. Initially a firm foundation is lied introducing the primal- respectively dual-problem and how the relation can be used to define optimization methods. The finale of those is ADMM. Its predecessors are also properly introduced in this thesis. Then using the same data as in prediction and analysis" [1], an experiment looking into parameter effects in both ADMM and Elastic Net is conducted. Some intuitions are confirmed, such as penalty effects and similar. Notably is the robustness of ADMM, despite this it can be more or less effective. Example wise it was found that some control parameters in ADMM and Elastic Net has a firm relations and should be properly chosen. [1] L. Bengtsson, "Late blight prediction and analysis", 2017. Student Paper.

  AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)