GPflow/GPflow
Gaussian processes in TensorFlow
What it solves
GPflow is a Python package designed to simplify the construction of Gaussian process (GP) models. It provides a framework for implementing modern Gaussian process inference, supporting composable kernels and likelihoods to handle complex data patterns.
How it works
Built on top of TensorFlow 2.4+ and TensorFlow Probability, GPflow leverages these libraries to perform the underlying computations. This architecture allows the models to execute quickly on GPUs, enabling more efficient processing of larger datasets.
Who it’s for
It is intended for researchers and developers who need to build and deploy Gaussian process models for tasks such as regression, Bayesian optimization, and molecular property prediction.
Highlights
- GPU Acceleration: Uses TensorFlow to enable fast execution on GPUs.
- Composable Components: Supports the creation of complex models through composable kernels and likelihoods.
- Extensible Framework: Serves as the foundation for other specialized projects like Trieste for Bayesian optimization.
- Modern Inference: Implements up-to-date Gaussian process inference techniques.