Model space-time data with Gaussian processes.
Geostat makes it easy to write Gaussian Process (GP) models with complex covariance functions. It uses maximum likelihood to fit model parameters. Under the hood it uses Tensorflow to fit models and do inference on GPUs. A good consumer GPU such as an Nvidia RTX 4090 can handle 10k data points.
Install Geostat using pip:
pip install geostat
- An introduction to Geostat. In Geostat, we create one model that is used to create synthetic data according to provided parameters, and we create a second model that does the inverse: it takes the data and infers the parameters.
- Structured covariance functions. Here we show how a progressively more complex covariance function fits data better than simpler ones.
- Making predictions in a shape. Geostat has utility functions to make it easier to work with shapes.
- Gaussian processes in Tensorflow. A tutorial on how to implement Gaussian processes in Tensorflow.