The Tool Landscape
Session Overview
- Python
- AutoML
- Deep Learning
- LLMs
Tools
In general, tools can be used across types of learning (supervised, unsupervised, and semisupervised)
In general, tools can be used across types of data (image, text, gridded, tabular)
In general, tools can be used across platforms (locally, on the cloud)
In general, most tools can be used for both deep and not-so-deep learning.
Python
- SciKit Learn
- Dask ML
- Darts Time Series
- Many Automated Ones
SciKit Learn
- Extensive package supporting all common ML algorithms
- Allows deep level customization and tuning
- Even has pipeline features
- Vastly Extensible
- Mapped almost one to one on Dask via Dask ML
AutoML
Automatically develop and try many, many ML models to find good models with minimal coding.
Paid Options
Deep Learning
- TensorFlow
- PyTorch
- Now under Linux Foundation
- PyTorch Lightning
- Keras
- High level deep learning API
- Excellent starting point
- Integrates seamlessly with Tidymodels and SciKit Learn
- Python and R
LLMs
Tools
In general, tools can be used across types of learning (supervised, unsupervised, and semisupervised)
In general, tools can be used across types of data (image, text, gridded, tabular)
In general, tools can be used across platforms (locally, on the cloud)
In general, most tools can be used for both deep and not-so-deep learning.
Earth System Data Science in the Cloud