Machine Learning for Psychologists
PSY 2085
Subject & Catalog Number
Course Information
Description
This is a full-on applied machine learning (ML) course designed for graduate students in psychology and related fields. The first half of this course focuses on supervised learning methods for prediction/classification. We elaborate on the basic principles of predictive modeling such as sample splits, cross-validation, and model evaluation (confusion matrix), as well as more advanced ideas like modeling imbalanced data and model-agnostic methods for interpreting black box models such as feature importance, accumulated local effects, SHAP values, etc. In terms of particular ML algorithms that will be covered before midterm, we start with basic logistic regression including penalization, followed by classification trees with important extensions such as (extreme gradient) boosting, bagging, and random forests, as well as support vector machines, basic neural networks, and ensemble learners. After midterm the first part is all about natural language processing (NLP) and deep learning. We start with basic text mining techniques such as sentiment analysis, topic models, and word embeddings. Then we cover more advanced transformer architectures like BERT for text-based modeling. The final block consists of two units on causal ML where we combine ML techniques with domain expertise for causal reasoning. We will use Python throughout this class.
Recommended Prep
The Psychology Department requires completion of PSY 1950 and PSY 1952 or equivalent courses before enrolling in this course. This course is intended for Harvard doctoral candidates in Psychology. Other students (e.g. Harvard undergraduates, Harvard graduate students from other fields, MIT graduate students) may obtain permission from the instructor to take the course.
Available for Harvard Cross Registration
NOTE: This course requires additional sections; you will be prompted to choose secondary components during the Add to Cart process
