Mid-Term Review

1 Species: Supervised Learning

1.1 Introduction

• Lecture: Introduction

  • What is Machine Learning?

    “Machine learning (ML) is the study of computer algorithms that can improve automatically through experience and by the use of data…” - Wikipedia

  • Supervised Learning vs Unsupervised Learning vs Semi-Supervised Learning vs Reinforcement Learning

  • ML workflow: split into training and test data, fit model on training data, predict on new data, evaluate model with test data

  • Accuracy vs precision

  • Artificial Intelligence (AI) > Machine Learning (ML) > Deep Learning (DL)

  • History of machine learning | Google Cloud

    • 1950: Alan Turing comes up with Turing test: originally called the “imitation game” (see also Imitation Game movie with Benedict Cumberbatch on cracking German codes to save WWII), which is the test of a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. He asked “Can machines think?” in his 1950 seminal paper “Computing Machinery and Intelligence.”

    • 2015: Google’s AlphaGo was the first program to best a professional player at Go, considered the most difficult board game in the world. With this defeat, computers officially beat human opponents in every classical board game.

  • Ecology and Environment applications:

    • individual > population > landscape > ecosystem > global

    • Deep Learning architecture of convolutional neural nets (CNNs in next module): sounds, images

  • Species Distribution Modeling

    • Elith & Leathwick (2009): geographic -> environmental (fit) -> geographic (predict)

    • modeling techniques: GLM, GAM, TREE (rpart, RandomForest)

    • Ecological Niche

      • Grinell: environmental conditions

      • Elton: env + biotic

      • Hutchinson: “n-dimensional hypervolume”

        • fundamental vs realized niche

• Lab: Species: explore

  • Query the Global Biodiversity Information Facility (GBIF.org) for observations

  • Fetch environmental data to be used as predictors

• Reading: Zhong et al. (2021) Machine Learning: New Ideas and Tools in Environmental Science and Engineering

  • inputs: tabular, image, graph, text

  • applications: making predictions; extracting feature importance; detecting anomalies; and discovering new materials/chemicals

  • workflow: preparation, model development, interpretation and deployment

1.2 Logistic Regression

• Lecture: Logistic Regression

• Lab: Species: regress

• Reading: Elith & Leathwick (2009) Species Distribution Models: Ecological Explanation and Prediction Across Space and Time

  • Key steps: gathering relevant data; assessing its adequacy (the accuracy and comprehensiveness of the species data; the relevance and completeness of the predictors); deciding how to deal with correlated predictor variables; selecting an appropriate modeling algorithm; fitting the model to the training data; evaluating the model including the realism of fitted response functions, the model’s fit to data, characteristics of residuals, and predictive performance on test data; mapping predictions to geographic space; selecting a threshold if continuous predictions need reduction to a binary map; and iterating the process to improve the model in light of knowledge gained throughout the process.

1.3 Decision Trees

• Lecture: Decision Trees

• Lab: Species: trees

• Reading: Evans et al. (2011) Modeling Species Distribution and Change Using Random Forest

  • issues: complex non-linear interactions, spatial autocorrelation, high-dimensionality, non-stationary, historic signal, anisotropy, and scale

  • Classification and Regression Trees versus Random Forest Algorithm (ensemble)

  • Model Selection (parsimony), Imbalanced Data, Model Validation

  • Current and Future Prediction

1.4 Model Evaluation

• Lecture: Model Evaluation

• Lab: Species: evaluate

• Reading: Andrade et al. (2020) ENMTML: An R package for a straightforward construction of complex ecological niche models

  • Overview of issues and methodologies

2 Communities: Unsupervised Learning

2.1 Clustering

• Lecture: Biodiversity; Clustering

• Lab: Communities: cluster

• Reading: Ch. 8-9 (p.123-152) of Kindt & Coe (2005)
  Ch. 8 Analysis of differences in species composition; Ch. 9 Analysis of ecological distance by clustering

  • Ecological Distance metrics to build matrices of between site dissimilarity based on species composition: Euclidean, Manhattan, Bray-Curtis

  • Hierarchical clustering in ecological context

  • Dendrograms

  • Complements ordination

2.2 Ordination

• Lecture: Ordination

• Lab: Communities: ordinate

3 Reserves: Optimization

3.1 Conservation Planning

• Lecture: Conservation Planning

• Lab: Reserves