2019 January 28¶
Location: Coors Tech¶
Attending:¶
Thomas Iga Jihyun Bane Rosie
Key takeaways¶
- Jihyun got an MSE of 1.5-ish in the Kaggle competition (top 200)
- Rosie is working on a cool 2D CNN senior design project for fault detection
- Jihyun will give a brief tutorial on boosting next week
Problems¶
- Rosie - getting started with fault detection
- Thomas - how can we scale LANL project features without precomputing them?
Comments¶
- Rosie is working with 3D migrated DAS images
- Her inputs are two different migrated images from active and passive sources
- Her output is a fault map/volume
- She could try a 2D CNN in Keras to start
- LANL Kaggle kernels precompute features before deriving scaling
Topics of discussion for next week¶
- Boosting tutorial from Jihyun