In this workshop, we’ll first discuss core machine learning concepts such as: choosing loss functions and evaluation metrics; splitting the data into training, validation, and testing sets; and cross-validation patterns for tuning hyper-parameters. Next, we’ll apply these concepts to train models for identifying isolated letters from speech (https://archive.ics.uci.edu/ml/datasets/isolet).

Specifically, we’ll apply the elastic net (a generalization of ridge and lasso regression), random forests, and gradient boosting to this task.  We’ll briefly discuss each model/method but our primary focus will be on understanding the core functionality of the related R packages (glmnet, randomForests, xgboost) and tuning associated hyper-parameters.