A Crash Course in Machine Learning Methods (3-Part Series)

Course Instructor: Olga Demler, PhD

Overview:

The Brigham Research Education Program is excited to once again offer “A Crash Course in Machine Learning Methods,” a 3-part course that will be taught virtually by Olga Demler, PhD over the course of three Fridays in November (4, 11, 18).

In this course we will review the Machine Learning methods used in medical research. We will also provide a broad overview of Deep Learning methods.

Course Objectives:

The goal of this course is to develop an intuition for each method and become familiar with the language used in this area. After introducing the basic concepts, we will offer an optional hands-on tutorial of applying these algorithms in the R programming language.

Target Audience:

This course assumes a working knowledge of R and intermediate statistical analysis, including linear and logistic regressions and linear discriminant analysis.

This course is based on recent developments in the field (references will be provided) and the books “The Elements of Statistical Learning” by Friedman, Hastie, Tibshirani (https://web.stanford.edu/~hastie/Papers/ESLII.pdf) and “An introduction to statistical learning, second edition” by James, Witten, Hastie, Tibshirani (https://www.statlearning.com/).

Syllabus:

Day 1 | Fri, Nov 4 | 9:00-11:30AM | Machine Learning Methods I

1. Broad Overview of Machine Learning and Deep Learning Methods
2. Supervised Learning Algorithms
   • Classification and Regression Trees
   • Random Forests
   • XGBoost
   • Support Vector Machines
3. Tutorial

Day 2 | Fri, Nov 11 | 9:00-11:30AM | Machine Learning Methods II

1. Supervised Learning Algorithms (cont.)
   • Elastic Net:
     • LASSO
     • Ridge regression
2. Unsupervised Methods
   • Dimension Reduction
     • PCA vs Factor Analysis
   • Classification and Pattern Recognition
     • K means clustering
     • tSNE and UMAP
3. Tutorial

Day 3 | Fri, Nov 18 | 9:00-11:30AM | Machine Learning Methods III

1. Broad Overview of Deep Learning Methods
   • Convolutional Neural Networks
2. Other Topics
   • Adjusting for Multiple Comparisons: controlling for family-wise type 1 error rate: Bonferroni; controlling for False Discovery Rate: FDR, permutation-based FDR
   • Cross-Validation
3. Tutorial

Summary:

• Date: Fridays November 4, 11, and 18
• Time: 9:00AM – 11:30AM
• Location: Virtual (Zoom)

The deadline to register is EOD Tues, Nov 1, 2022.