Machine Learning Training for Natural Hazards Engineering 2022

Welcome to the NHERI SimCenter and DesignSafe practicum for machine learning (ML), emphasizing ML’s applications to Natural Hazards Engineering. The practicum is aimed toward engineering students and faculty interested in enhancing their research by applying ML to natural hazards engineering problems. Through instruction and exercises, the training will provide you with fundamentals and these fundamentals will be reinforced with hands-on examples so that you gain confidence in utilizing ML in your work. As a basis, the training will introduce opportunities and challenges presented in Earthquake Engineering, cover explainability and deriving physics using AI, and introduce the ML techniques that are used by SimCenter for generating asset inventories that can be used in regional hazard simulation. Participants will learn salient ML techniques behind image classification and semantic segmentation, and utilize the SimCenter’s BRAILS framework.

Prerequisites: Some Python programming knowledge is required. Participants w ith no or little Python experience should review the Python quick start material (Chapters 1-6) from SimCenter Programming Bootcamp 2022

Contents:

• Presenters
• Syllabus
  • Day 1: ML in Earthquake Engineering: Opportunities and Challenges
  • Day 2: Explainable AI in Earthquake Engineering
  • Day 3: Image Classification
  • Day 4: Semantic Segmentation
  • Day 5: Inventory Generation with BRAILS
• Video Content
  • Day 1: Machine Learning in Infrastructure Risk and Resilience Assessment - A Constructive Critique
  • Day 2: Explainable AI in Earthquake Engineering
  • Day 3: Neural Network Architectures
  • Day 4: Semantic Segmentation
  • Day 5: Inventory Generation with BRAILS
• Notebooks
  • Day 3: Image Classification
  • Day 4: Semantic Segmentation
  • Day 5: Inventory Generation with BRAILS

Note

1. The latter half of the workshop will consist of pre-recorded training videos and daily, live, online, hands-on exercises. Before each live exercise session, participants are required to watch the short instructional videos when available. Participants who do not watch the instructional videos will not be prepared for the live exercises.

2. Course material is available through the NHERI-SimCenter/SimCenter_DesignSafe_ML_2022 repository on GitHub

3. You will be using Jupyter Notebooks on DesignSafe to perform the exercises.