Machine Learning for Process Analytical Technology (PAT)

April 22, 2024

New Brunswick, NJ

Machine Learning for PAT is aimed at spectroscopists and other scientists and engineers who want to be able to use machine learning methods to develop their own linear and non-linear models for calibration/regression or sample classification. The course addresses linear methods briefly but will focus on machine learning methods designed to deal with non-linear data. Emphasis will be on applying these techniques to spectroscopic, especially NIR and Raman data which have become ubiquitous in the process environment. Methods for instrument calibration, sample classification and exploratory data analysis will be covered.

About the Instructors

The course will be led by Eigenvector Research President and PLS_Toolbox creator Barry M. Wise, Ph.D. Dr. Wise has delivered over 200 chemometrics courses at scientific conferences, on-site for companies, on-line and at open sites. Wise is winner of the 2001 Eastern Analytical Symposium Award for Achievements in Chemometrics and the Herman Wold Medal in Gold for his dedication to teaching. Eigenvector Senior Data Scientist and Principal Consultant Manuel A. Palacios, Ph.D. will also present. Dr. Palacios has extensive experience with linear and non-linear modeling methods having successfully developed and deployed hundreds of models for calibration and classification.

Prerequisites

It is recommended that participants be familiar with Principal Components Analysis (PCA) and multivariate regression methods such as Partial Least Squares (PLS).

Course Outline

Classification and Regression Methods

Linear and Non-linear Data

Partial Least Squares (PLS) Regression

Partial Least Squares Discriminant Analysis (PLS-DA)

Hierarchical Modeling (HMB and HMAC)

Locally Weighted Regression (LWR)

Soft Independent Modeling of Class Analogy (SIMCA)

Support Vector Machine Discriminant Analysis (SVMDA)

Artificial Neural Networks Regression and Discriminant Analysis (ANN, ANNDA)

Boosted Regression Trees Discriminant Analysis (XGBoostDA)

Conclusion/ Choosing Which Method

Data Compression for Visualization and Preprocessing

Brief review of Principal Components Analysis (PCA)

t-distributed Stochastic Neighbor Embedding (t-SNE)

Uniform Manifold Approximation and Projection (UMAP)

Conclusion/ Choosing Which Method

Discussion and Questions

Registration and Venue

Machine Learning for PAT is offered prior to the APACT USA Conference in New Brunswick, NJ USA, April 23-25, 2024. The course will be held at the Hyatt Regency New Brunswick. Registration is done through the conference website.