UMDataScienceLab / Consensus_Bayesian_Opt

Consensus_Bayesian_Opt

Code for the paper entitled "Collaborative and Distributed Bayesian Optimization via Consensus: Showcasing the Power of Collaboration for Optimal Design"

Our Algorithm

Collaborative and Distributed Bayesian Optimization via Consensus - CBOC

Requirement

Python version 3.9.16

Code files

We here provide three examples, and users are free to test our algorithm using other functions by modifying our code.

(1) Branin_code.ipynb: code to test the collaborative algorithm on the Branin function.

(2) Shekel_code.ipynb: code to test the collaborative algorithm on the Shekel function.

(3) Levy_code.ipynb: code to test the collaborative algorithm on the Levy function.

Modification

Our CBOC algorithm can also be applied to many other testing functions. Please modify the code to run more simulations. For example, in the Levy_code.ipynb file:

• Change "Levy()" in the line "globals()['original_train_Y_%s' % i] = globals()['temp_train_Y_%s' % i] = globals()['train_Y_%s' % i] = Levy(dim=input_dim, negate=True)(eval('train_X_%d' % (i))).unsqueeze(-1)" to other function name.
• When using other testing functions, please also assign new values to global variables such as "num_dev", "input_dim", etc.