1/20/2024 0 Comments Instaling abcDBThe single-fit function fit_one_model, on which the list-fitting functions above are built (so that a user may implement their preferred parallelisation) The model_search_rounds function to continue a search from where another left-offįunctions to be used as model_list_fitter argument: fit_mods_not_parallel and fit_mods_parallel_processes (using multiprocessing’s Pool) Main model search function plus a few convenient tools (refer to the section below for context): The directly importable elements from this package are essentially those required to customise any of the arguments of the Selection of output from the above example if the original data was subsampled) #fit_kex(other_X, other_Y, best_mods.kernel_expression, 10) show () # Not required for plotly = True above # Later fit one of the top kernels to new data (e.g. array ()), ' \n ' ) from matplotlib import pyplot as plt plt. join () + f ' \n ' ) print ( ' \n\n Top-3 models \' details:' ) for bm in best_mods : model_printout ( bm, plotly = False ) # See the definition of this convenience function for examples of model details' extraction print ( 'Prediction at X = 11:', bm. randn ( 101, 1 ) # Main function call with default arguments best_mods, all_mods, all_exprs, expanded, not_expanded = explore_model_space ( X, Y, start_kernels = start_kernels, p_rules = production_rules, utility_function = BIC, rounds = 2, beam =, restarts = 5, model_list_fitter = fit_mods_parallel_processes, optimiser = GPy_optimisers, max_retries = 1, verbose = True ) print ( ' \n Full lists of models by round:' ) for mod_depth in all_mods : print ( ', '. The function should be called from within a if _name_ = '_main_': for full OS-agnostic use.Ī minimal example to showcase the various parameters follows: import numpy as np from GPy_ABCD import * if _name_ = '_main_' : # Example data X = np. Note that with the default model_list_fitter = fit_mods_parallel_processes argument See its description for parameter information and type hints. The main function exported by this package is explore_model_space GPy-ABCD: A Configurable Automatic Bayesian Covariance Discovery Implementation.Ĩth ICML Workshop on Automated Machine Learning (2021) Installation pip install GPy_ABCD Usage See the picture in Usage below for an example input/output and read the paper for further details:įletcher, T., Bundy, A., & Nuamah, K. Of its top few results’ kernels (on the full dataset if subsampled before). (possibly on subsampled datasets for efficiency), then followed by more direct exploration The usefulness of ABCD is in identifying the underlying shape of data, but the process isĬomputationally expensive, therefore a typical use for it is in initial data overviews covariances) constructed by iteratively combining a small set of simple ones,Īnd returning the best fitting models using them ĭue to its modularity, it is capable of generating simple text descriptions of theįits based on the identified functional shapes. GPy-ABCD is a basic implementation with GPy of an Automatic Bayesian Covariance Discovery (ABCD) system.īriefly, ABCD is a (Gaussian Process) modelling method which consists in exploring a space of modular kernels (Temporary note: the “failing” build badge above is due to the workflow pip not finding GPy 1.12.0 for some reason the tests are successful)
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |