Featured
Dragonfly is a Python library for scalable Bayesian optimisation.
It provides an array of tools to scale up Bayesian optimisation to expensive large
scale problems including features/functionality for high
dimensional optimisation, parallel
evaluations in synchronous or asynchronous settings, multi-fidelity optimisation,
and multi-objective optimisation.
Citation:
Tuning Hyperparameters without Grad Students: Scalable and Robust Bayesian
Optimisation with Dragonfl
,
Arxiv 2019.
A matlab implementation of additive upper confidence bound methods in
Gaussian process bandits.
The additive model is especially useful for
high dimensional Bayesian Optimisation.
Citation:
High Dimensional Bayesian Optimisation and Bandits via Additive Models
,
ICML 2015.
Dragonfly is a Python library for scalable Bayesian optimisation.
It provides an array of tools to scale up Bayesian optimisation to expensive large
scale problems including features/functionality for high
dimensional optimisation, parallel
evaluations in synchronous or asynchronous settings, multi-fidelity optimisation,
and multi-objective optimisation.
Citation:
Tuning Hyperparameters without Grad Students: Scalable and Robust Bayesian
Optimisation with Dragonfl
,
Arxiv 2019.
A python implementation of parallelised Bayesian optimisation using Thompson sampling.
We provide implementations of both the synchronous and asynchronous versions along
with experimental set ups in synthetic settings where the evaluation time is
modeled as a random variable.
Citation:
Parallelised Bayesian Optimisation via Thompson Sampling
,
AISTATS 2018.
A matlab library for estimating various information theoretic quantities such
as the Shannon/Renyi/Tsallis entropies, divergences, mutual informations and
their conditional versions.
Our estimators are also available in the ITE Toolbox.
Citation:
Nonparametric Von Mises Estimators for Entropies, Divergences and
Mutual Informations
,
NIPS 2015.
A matlab library for locally polynomial regression methods (including
Nadaraya Watson) using the Gaussian and Legendre kernels.
Citation:
Additive Approximations in High Dimensional Nonparametric Regression
via the SALSA
,
ICML 2016.
A matlab implementation of multi-fidelity Bayesian optimisation using Gaussian
processes.
Citation:
Gaussian Process Bandit Optimisation with Multi-fidelity Evaluations
,
NIPS 2016.
Python implementation of MPS (Myopic Posterior Sampling) for adaptive goal oriented
design of experiments (DoE). MPS is a general and flexible framework for adaptive
DoE where a practitioner may specify her goal via a reward function.
In encompasses a variety of existing adaptive DoE settings (e.g. active learning,
optimisation, level set estimation) but also enables DoE in application specific
settings where prior methods are not applicable.
Citation:
Myopic Posterior Sampling for Adaptive Goal Oriented Design of Experiments
.
Python implementation of NASBOT: Neural Architecture Search with Bayesian Optimisation
and Optimal Transport. NASBOT is a method for tuning convolutional neural networks
and multi-layer perceptrions using Gaussian processes and optimal transport.
Some datasets are available here.
Citation:
Neural Architecture Search with Bayesian Optimisation and Optimal
Transport.
A matlab library for estimating nonparametric hidden Markov models.
We use
Chebfun
to compute functional SVDs using Chebyshev polynomials.
Citation:
Estimating HMMs with Nonparametric Emissions via Spectral Decompositions
of Continuous Matrices
,
NIPS 2016.
A matlab implementation of kernel ridge regression using additive kernels.
Citation:
Additive Approximations in High Dimensional Nonparametric Regression
via the SALSA
,
ICML 2016.