GPflow: A Gaussian Process Library using TensorFlow

By Alexander G. de G. Matthews, Mark van der Wilk, T. Nickson, Keisuke Fujii, Alexis Boukouvalas, Pablo León-Villagrá, Zoubin Ghahramani, James Hensman. 2016

750 citations recorded by the source corpus.

GPflow is a Gaussian process library that uses TensorFlow for its core computations and Python for its front end. The distinguishing features of GPflow are that it uses variational inference as the primary approximation method, provides concise code through the use of automatic differentiation, has been engineered with a particular emphasis on software testing and is able to exploit GPU hardware.

Subfields: Artificial Intelligence, Probabilistic & Bayesian Methods, ML Systems & Hardware

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Probabilistic & Bayesian Methods

In ML Systems & Hardware

Citations in the corpus

Functional Regularisation for Continual Learning using Gaussian Processes (2019, methodology): “The FRCL methods have been implemented using GPflow (Matthews et al., 2017) where multi-class problems make use of the robust likelihood.”
Deep Bayesian Bandits Showdown: An Empirical Comparison of Bayesian Deep Networks for Thompson Sampling (2018, methodology): “For the sparse version we used a Sparse Variational GP (Hensman et al., 2015) with the same kernel and with 300 inducing points, trained via minibatch stochastic gradient descent (Matthews et al., 2017).”
Adversarial Examples, Uncertainty, and Transfer Testing Robustness in Gaussian Process Hybrid Deep Networks (2017, methodology): “We first describe in the next section the structure of these GPDNN models as well as indicating how they can be trained end-to-end in a scalable manner using variational methods, GPflow [39] and TensorFlow [2].”
Safe Model-based Reinforcement Learning with Stability Guarantees (2017, methodology): “Experiments A Python implementation of Algorithm 1 and the experiments based on TensorFlow [37] and GPflow [38] is available at https://github.”
Streaming Sparse Gaussian Process Approximations (2017, methodology): “The method was implemented on GPflow [20] and compared against GPflow’s version of the following baselines: exact GP (GP), sparse GP using the collapsed bound (SGP), and stochastic variational inference using the uncollapsed bound (SVI).”
Gaussian Process Behaviour in Wide Deep Neural Networks (2018, methodology): “This will be facilitated by the release of our covariance function code built on GPflow (Matthews et al., 2017).”
Few-shot learning of neural networks from scratch by pseudo example optimization (2018, methodology): “We embedded 2https://github.com/GPflow/GPflow/tree/master/doc/source/notebooks/data each of two-dimensional examples onto a 100-dimensional vector space by a linear transformation, where a transformation matrix was randomly determined in advance, and we selected five supervised training examples per”
Deep convolutional Gaussian processes (2018, methodology): “, 2016) implementation is compatible with the GPow framework (Mahews et al., 2017) and freely available online 1.”
Auto-Differentiating Linear Algebra (2017, neutral): “The GPFlow project [6] (code at https://github.”
Bayesian Layers: A Module for Neural Network Uncertainty (2018, methodology): “Alternatively, the software may use high-level abstractions in order to specify and fit specific model classes with a hand-derived algorithm (GPy, 2012; Vanhatalo et al., 2013; Matthews et al., 2017).”
Efficient Gaussian Process Classification Using Polya-Gamma Data Augmentation (2018, methodology): “We compare our proposed method, efficient Gaussian process classification (X-GPC), with the state-of-the-art methods SVGPC (Salimbeni, Eleftheriadis, and Hensman 2018), provided in the package GPflow3 (Matthews et al. 2017), which builds on TensorFlow and the EP approach EPGPC by Hernández-Lobato an”
Partitioned Variational Inference: A unified framework encompassing federated and continual learning (2018, methodology): “While this does not change the fixed-point solution compared to the PVI algorithm, it allows existing sparse global VI implementations such as that in GPflow [Matthews et al., 2017] to be easily extended and deployed.”
GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration (2018, methodology): “For Exact and SGPR, we compare BBMM against Cholesky-based inference engines implemented in GPFlow [33].”
Imitation networks: Few-shot learning of neural networks from scratch (2018, positive): “We used GPflow [Matthews et al., 2017] with variational Gaussian approximation [Opper and Archambeau, 2009] and L-BFGSB optimization [Zhu et al., 1997] for GP inference, and utilized mean predictions as pseudo supervisions.”