🕘 Meeting notes: Machine Learning WG
- AMS Annual is virtual. Abstract Deadline August 24, AI conference. Relevant sessions:
- aI for climate applications
- blending ai with wx and climate models
- open datasets for ai research and applications
- python frameworks for reproducible communicable
- using hpc to accelerate ML
- talk today at 3 EDT, by Jon Weyn (UW) “S2S forecasting using large ensembles of data-driven global weather prediction models”
- Thursdays NOAA STAR sessions:
- KDD (virtual august 23-27)
- Deep Learning for Science School: https://dl4sci-school.lbl.gov/
- How to make pangeo work better for ML?
- need example
- Tutorials through pangeo gallery, binderhubs
- job submission
- kubectl apply jobs
- AWS Batch
- Google AI Jobs
- Genomics Pipeline thing
- social engineering, get google / aws to install xarray and zarr into their ML images
- connect to sagemaker?
- Instructions on how to use their cloud provider's ML resources
- Easily spin up / down GPUs to minimize costs
- Need more comments on workflow/infrastructure examples:
- Put out survey to see what the community wants / what they are using / what do they want to use?
- any workshops teach cloud authentication?
- Survey? Google Forms
- Focus: the hard (non-scientifically interesting) stuff
@Laique Merlin Djeutchouang Univ Cape Town, Ocean + ML
- Summer school.
- huge live stream turnout (1000s!)
- Hackathon (300 participants, groups of 5)
- challenge problems
- jupyter notebook
- Google collab + AWS Jupyterhub