HN Explorer

Ask HN: A good course for spatial Machine Learning

1 points by anakaine ↗ HN
There are some excellent primer and hands on courses for machine learning from the likes of deeplearning.ai and fast.ai. Neither of these courses, however, offer anything with regards to deep learning for spatial data.

Whilst there are some basic overviews from industry heavyweight ESRI, the ESRI tools rely on high cost paid licensing to operate, meaning high cost of learning.

So, HN bretheren, have you found any good resources for learning and understanding ML as applied to spatial data, preferably with the sort of interactive student delivery and course quality of the above? Could be desktop GIS based, Python, Jupyter Notebooks with Pytorch, TensorFlow, etc. Options are good.

2 comments

[ 0.20 ms ] story [ 16.5 ms ] thread
p1esk
Take any ML course to learn the basics, then apply what you learned to process your data. Don’t fret over course quality or “student delivery” - pick any resource and start learning.
anakaine
I guess my concern is that spatial data and its interactions are quite distinct from traditional tabular data, or imagery data. I'm less interested in learning how to classify trees and roofs from a satellite image, and more in the interaction of various layers of differing types.

That said, I could probably adjust inputs so that there is a tabular training set. Gut feel is that there will still be some big holes in my knowledge by travelling a traditional ML route.