Artificial Intelligence for Computational Sustainability: A Lab Companion/Machine Learning for Prediction/Instructor Summary for Species Distribution Modeling Lab
Jump to navigation
Jump to search
Instructor Summary for Species Distribution Modeling lab[edit | edit source]
Keywords | Description |
---|---|
Summary | In this lab, students use maximum entropy modeling to predict the distribution of various tree species based on a set of climate variables. It explores single variable modeling, multivariate modeling, examining differences due to climate change, and includes a brief discussion of ROC analysis. The lab uses the free Maxent software (written in Java) developed by Robert Schapire's group at Princeton, and includes instructions to allow students to download, install, and use the software without instructor or system administrator support. |
AI Topics | maximum entropy modeling, learning from positive examples, ROC analysis |
Sustainability Topics | species distribution modeling |
Audience | Little AI or CS knowledge is assumed, so this is suitable for Introductory AI courses, or even general education courses. With adaptation, it could be used in CS for non-majors. |
Difficulty | The lab takes students step-by-step through using the software and performing the analysis. It is estimated to take approximately 4-5 hours, and can easily be broken up along the section boundaries. |
Strengths | It provides a great introduction to the application of machine learning to species distribution modeling based on presence data. It assumes little AI knowledge, and therefore can be used in a wide variety of courses. However, it can also be used to support in-depth discussions of Maxent, and can therefore be used in Introductory or Advanced AI courses. |
Weaknesses | In its current form, students simply use the Maxent software and draw conclusions from the resulting figures. To be more useful in advanced AI courses, it needs to be augmented with either in-depth discussion of Maxent, or additional questions that involve in-depth analysis or implementation. |
Dependencies | Little AI or CS knowledge is assumed, but this assignment can easily be used to augment an in-depth discussion of Maxent. The Maxent software is in Java (in a single jar file) and is therefore cross-platform. Students can download, install, and run the software without instructor or system administrator support. |
Variants | This lab can be used to augment discussions of Maxent, where students read the papers on Maxent and go into the details of the algorithm, and then apply the algorithm to learn species distribution models. This lab can also be used to augment discussions on ROC analysis, since ROC curves are produced for each model. |