What is autotuning? In this SAS How To Tutorial, SAS’ Robert Blanchard explains that “autotuning is a technology developed by SAS to intelligently and efficiently search a decision space.” Data scientists enter this “decision space” when choosing how to define a base architecture of their model or code base.
Robert explains three aspects of autotuning in a deep learning context: globalized search, localized search and an in parallel method using SAS - specifically with the Black-Box Optimization Solver (or Solve Black-Box). You’ll see how to use Solve Black-Box to find a good model architecture or customize a backbone model to better fit your data.
Chapters
00:00 - Intro
00:56 - Autotuning in a deep learning context
02:25 - Autotuning overview
03:44 - Globalized search through a genetic algorithm
06:08 - Localized search method
07:17 - Demo: How to perform autotuning for deep learning in SAS
08:44 - Build a customized model in SAS
13:40 - Explanation of the Solve Black-Box code
Learn more about SAS Software
◉ What is deep learning? - www.sas.com/en_us/insights/an...
◉ Data Mining and Machine Learning Community - communities.sas.com/t5/SAS-Da...
◉ Black-Box Optimization Solver (Solve Black-Box) -documentation.sas.com/doc/en/...
◉ AI & Machine Learning Professional Training - www.sas.com/en_us/training/ac...
Learn more about deep learning from Robert
◉ • Learn SAS with Robert ...
SUBSCRIBE TO THE SAS USERS RU-vid CHANNEL #SASUsers #LearnSAS
ru-vid.com?sub_...
ABOUT SAS
SAS is a trusted analytics powerhouse for organizations seeking immediate value from their data. A deep bench of analytics solutions and broad industry knowledge keep our customers coming back and feeling confident. With SAS®, you can discover insights from your data and make sense of it all. Identify what’s working and fix what isn’t. Make more intelligent decisions. And drive relevant change.
CONNECT WITH SAS
SAS ► www.sas.com
SAS Customer Support ► support.sas.com
SAS Communities ► communities.sas.com
SAS Analytics Explorers ► explorers.sas.com
Facebook ► / sassoftware
Twitter ► / sassoftware
LinkedIn ► / sas
Blogs ► blogs.sas.com
RSS ► www.sas.com/rss
30 июл 2024