CCP | An introduction to modeling Naive Discriminative Learning
An introduction to modeling Naive Discriminative Learning

Antti Arppe, University of Alberta (Canada)

This course provides a practical introduction to applying the Naïve Discriminative Learning model (NDL) on linguistic data using the ndl package in the R statistical programming environment.

Naive Discriminative Learning (Baayen, Milin, Filipovic, Durdevic, Hendrix & Marelli 2011) models an incremental learning process of distinguishing alternative categorical outcomes, through experiencing piecemeal their co-occurrences with contextual cues. NDL is based on the Rescorla-Wagner (1972) equations, which have proven to be surprisingly fruitful in human and animal learning, and Danks’ (2003) solution for estimating the end result of such a learning process. Compared to standard classification algorithms, NDL may overfit the data, albeit gracefully.

The course combines short lectures with practical hands-on sessions using the ndl package (Shaoul, Arppe, Hendrix, Milin & Baayen 2014) with a number of example datasets and comparing the results with machine learning methods (e.g. Generalized Linear Models) in the R statistical programming environment. Participants are free to bring their own data for applying the ndl package. Basic knowledge of statistics and using R are assumed.