Open discussions on specific topics selected by the Software Working Group and selected from the list of SWG Topics For Discussion.
Tuesday, March 22, 2022 - Hands-on Machine Learning Study Materials (Beginner to Intermediate Levels) for Research Software Engineers: An attempt to break down ML.
We will look at some core ideas of ML and work out a geospatial image classification problem. Towards the end we will ponder on some questions : What kind of problems should be solved by an Ai-first approach? Is AI a black-box? How to put AI into production? moderated by Minu Mathew
Presented Jupyter notebook with specific Deep Learning example - Planet Dataset Image Classification using Kafka
Recording:https://uofi.box.com/s/6u7oegm6lvyrie4tw1ndnvickginxmb3
Attendees:
Sandeep Puthanveetil Satheesan
Discussion:
- Sandeep Puthanveetil Satheesan What kind of resources can we provide to staff? Minu showed a slide with a list of links.
- Vismayak Mohanarajan How important is it to know the inner workings of a specific method?
- Minu Mathew It helps, even if you don't know all the inner workings, it's always helpful to understand the bigger picture.
- Sandeep Puthanveetil Satheesan if you are using existing models, you need to just know t
- Luigi Marini How to we create a guide to learn how to pick a method and learn the most common algorithm within that method?
- Kastan Day Understand structured vs structured
- Santiago Nunez-Corrales Start with the easiest tool first, occams razor
- Sandeep Puthanveetil Satheesan Have quick iterations with simpler models and setup your pipeline
- Vismayak Mohanarajan one of the aims of the focus groups could be a flow chart to help people know how to select specific models
- Kastan Day pretrained models are really helpful. Also dev MLOps approaches, for example https://wandb.ai/site
- Sandeep Puthanveetil Satheesan Focus will get started next. Please join! Hands-on Machine Learning Study Materials for Research Software Engineers
If you are interested in contributing to a Round Table, please see these links: