Description
Workshop Objective:
Introduce the theory and practical applications of machine learning within the Energy industry as means to help improve data-driven decision making.
The course will be delivered by renowned Geostatistics and Reservoir Modeling expert, Dr. Michael Pyrcz, a leader at daytum and an Associate Professor from University of Texas at Austin, Hildebrand Department of Petroleum Engineering and Geosystems Engineering
Course materials will consist of web-hosted lecture materials, instructor led, well-documented methods, workflow demonstrations, and coding examples. The workshop will be delivered via a cloud-based Jupyter Lab session, so the only thing you need to bring are your laptop and your fingers!
Registrants will be required to provide their github username (please include it in the comments section during the registration). We will be using github usernames to provide access to the Jupyter Lab which contains classroom notes and coding exercises.If you don't have a github account, please create a free one using this link: https://github.com/join?source=header-home.
Upon completion, all participants will be given SPE GCS Professional Develop Hours and Daytum Course Completion Certificate!
Detailed outline of the Workshop:
|
Time |
Topic |
Objectives |
|
8:00 AM - 8:30 AM |
Registration/Welcome |
|
|
8:30 AM - 9:30 AM |
Introduction to Machine Learning |
Provide definitions, fundamental concepts of inference and prediction along with the opportunity and limitations of machine learning. |
|
9:30 AM - 10:30 AM |
Inference: Dimensionality Reduction and Clustering |
Motivation and methods for inferential machine leaning methods including dimensionality reduction and clustering with potential applications |
|
10:30 AM - 11:30 AM |
Prediction: k Nearest Neighbors |
Motivation and methods for predictive machine learning methods including hyper parameter tuning with k nearest neighbors. |
|
11:30 AM - 12:30 PM |
Lunch |
|
|
12:30 PM - 1:30 PM |
Prediction: Tree based Regression |
Introduce tree-based modeling as one of the most interpretable machine learning prediction methods and a prerequisite for more powerful ensemble methods. |
|
1:30 PM - 2:30 PM |
Prediction: Ensemble Tree-based Regression |
Introduce predictive machine learning with ensemble methods for prediction with bagging and random forest. |
|
2:30 PM - 3:30 PM |
Prediction: Neural Networks |
Introduce and demonstrate neural networks as powerful and flexible machine learning prediction methods. |
|
3:30 PM - 4:30 PM |
Examples and Closing Remarks |
Summary of opportunities for integrating data analytics, statistics, machine learning and physics-based modelling to solve practical energy problems |