Asynchronous ϵ-Greedy Bayesian Optimisation

This repository contains the Python3 code for the experiments and results presented in:

George De Ath, Richard M. Everson, and Jonathan E. Fieldsend. Asynchronous ϵ-Greedy Bayesian Optimisation. Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence, PMLR 161:578-588, 2021.
Paper: https://proceedings.mlr.press/v161/de-ath21a.html

The repository also contains all training data used for the initialisation of the optimisation runs carried out, the optimisation results of each of the runs, and jupyter notebooks to generate the results, figures and tables in the paper.

If you would like help running experiments or just have questions about how the code works, please open an issue and we will do our best to offer help and advice.

Citation

If you use any part of this code in your work, please cite:

@inproceedings{death:aegis:2021,
    title={Asynchronous ϵ-Greedy Bayesian Optimisation},
    author = {George {De Ath} and Richard M. Everson and Jonathan E. Fieldsend},
    year = {2021},
    booktitle = {Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence},
    pages = {578--588},
    year = {2021},
    editor = {de Campos, Cassio and Maathuis, Marloes H.},
    volume = {161},
    series = {Proceedings of Machine Learning Research},
    month = {27--30 Jul},
}

Installation Instructions

> git clone https://github.com/georgedeath/aegis
> cd aegis
> conda create -n aegis python=3.7 numpy matplotlib scipy statsmodels tqdm docopt
> conda activate aegis
> conda install pytorch cpuonly botorch pytorch -c pytorch -c gpytorch
> pip install pyDOE2 pygmo==2.13.0 pygame box2d-py

• Profet benchmark (optional):

> pip install pybnn
> pip install git+https://github.com/amzn/emukit

Reproduction of experiments

The python file run_exp.py provides a convenient way to reproduce an individual experimental evaluation carried out the paper. It has the following syntax:

> conda activate aegis
> python run_exp.py --help
Usage:
    run_exp.py <problem_name> <run_no> <budget> <n_workers> <method> <time_function>
                [--problem_dim=<d>] [--epsilon=<e>] [--eta=<n>]
                [--acq_name=<acq_name>] [--repeat_no=<repno>]

Arguments:
    <problem_name>  Name of the problem to be optimised.
    <run_no>        Optimisation run number / Problem instance.
    <budget>        Number of function evaluations (including training data)
                    to carry out.
    <n_workers>     Number of asynchronous workers to run.
    <method>        Acquisition function name, choose from one of the list below.
    <time_function> Distribution to sample job times from.
                    Valid options: [halfnorm, pareto, exponential].

Acquisition functions:
    BatchTS [Thompson Sampling]
    aegisRandom (~) [AEGiS-RS]
    aegisParetoFront (~) [AEGiS]
    HalluBatchBO (*) [Kriging Believer]
    LocalPenalisationBatchBO (*) [Local Penalisation]
    HardLocalPenalisationBatchBO (*) [PLAyBOOK]
    (~) = Need to specify epsilon value.
    (*) = Need to specify a sequential acquisition function to use.

Options:
    --problem_dim=<d>   Specify problem dimensionality. Only needed when you're
                        optimising problems with a different dimensionality
                        than the default.
    --epsilon=<e>   Epsilon value for the aegis methods (~).
    --eta=<n>   Eta value for the aegis methods (~).
    --acq_name=<acq_name>   Sequential acquisition function name, only
                            used with methods marked with (*) above.
    --repeat_no=<repno> Used for the instance-based methods. Specifies the
                        run number of the instance. Note that each combination
                        of run_no and repeat_no will need to have its own
                        training data.

Reproduction of figures and tables in the paper

• AEGiS_results.ipynb contains the code to load and process the optimisation results (stored in results directory), as well as the code to produce all results figures and tables used in the paper and supplementary material.
• presentation_plots.ipynb contains the code to create the figures used in the presentation at UAI 2021.

Training data

The initial training locations for each of the 51 sets of Latin hypercube samples for the various noise levels are located in the data directory. The files are named like ProblemName_number.pt, e.g. first set of training locations for the Branin problem is stored in Branin_001.pt. Each of these files is a compressed numpy file created with torch.save. It has two torch.tensor arrays (Xtr and Ytr) containing the 2*D initial locations and their corresponding fitness values. Note that for problems that have a non-default dimensionality (e.g. Ackley with d=5), then the data files have the dimensionality appended, e.g. Ackley5_001.pt; see the suite of available synthetic test problems. To load and inspect the training data, use the following instructions:

> python
>>> import torch
>>> data = torch.load('data/Ackley5_001.pt')
>>> Xtr = data['Xtr']  # Training data locations
>>> Ytr = data['Ytr']  # Corresponding function values
>>> Xtr.shape, Ytr.shape
(torch.Size([10, 5]), torch.Size([10]))

Optimisation results

The results of all optimisation runs can be found in the results directory. The notebook AEGiS_results.ipynb shows how to a load all the experimental data and to subsequently plot it.