Neural Networks: Essentials

This course covers the essential components of a neural network architecture and details how to modify the structure as needed. Modifying the neural network’s error functions, activation functions, and hidden units is discussed. You learn how to programmatically build a neural network, implement early stopping, build autoencoders for a predictive model, and perform an intelligent search of the model hyperparameter values. The course finishes with an introduction to deep learning. The NEURAL procedure from SAS 9.4 and the NNET procedure from SAS Viya are used throughout the course.

Learn how to

• programmatically build neural networks in SAS
• modify neural network structure for better performance
• enhance data with autoencoders and synthetic observations.

Who should attend

Those interested in learning how to program neural networks in SAS

Formats available	Duration
Classroom:	2.0 days
Live Web Classroom:	3 half-day session(s)
e-Learning:	10.5 hours/180 day license

Before taking this course, you should have the following:

• some familiarity with SAS and SQL programming
• an understanding of predictive modeling
• a basic understanding of calculus.

This course addresses SAS Viya software.

Neural Networks: Essentials

• introduction
• multilayer perceptrons
• neural network modeling paradigm
• other considerations
• using a surrogate model to interpret neural network predictions

Neural Networks: Details

• parameter estimation
• numerical optimization methods
• regularization
• unbalanced data
• SAS search optimizations (self-study)

Tuning a Neural Network

• selecting hyperparameters with autotuning

Introduction to Deep Learning

• introduction to deep learning
• autoencoders

Radial Basis Function Networks (Self-Study)