Code and data for BFOR 516 Fall 2022
Algorithm - In mathematics and computer science, an algorithm is a finite sequence of rigorous instructions, typically used to solve a class of specific problems or to perform a computation. (Wikipedia).
Examples: Multi-layer Perceptron, Naive Bayes, Stochastic Gradient Descent
Model - "Machine learning models are output by algorithms and are comprised of model data and a prediction algorithm." (Article). You can think of the model as a map between the predictors and an outcome.
Other Names: estimator (sklearn terminology)
Predictor - A predictor is data that may be related to an outcome.
Other names: attribute, column, feature, independent variable.
Parameter (hyperparamter) - Before you train a model, you must specify
parameters that affect how the model is fit. When creating a model with
sklearn, the parameters are set before training and affect how the
algorithm learns. Generally, these are known as hyperparamters.
Examples: Maximum depth in a decision tree or hidden layers in a neural network.
Parameter (model) - These are the values that are set and updated inside a model. Example: In a logistic regression, coefficients are paramters that are learned during the fitting process. (Guide explaining the difference between hyperparameters and parameters).