Goldenberry

Abstract

Orange is an open-source component-based software framework, featuring visual and scripting interfaces for many machine learning algorithms. Currently it does not support Estimation of Distribution Algorithms (EDA) or other methods for black-box optimization. Here we introduce Goldenberry, an Orange toolbox of EDA visual components for stochastic search-based optimization. Its main purpose is to provide an user-friendly workbench for researchers and practitioners, building upon the versatile visual front-end of Orange, and the powerful reuse and glue principles of componentbased software development. Architecture of the toolbox and implementation details are given, including description and working examples for the components included in its first release: cGA, UMDA, PBIL, TILDA, UMDAc, PBILc, BMDA, CostFunctionBuilder and BlackBoxTester.Goldenberry is open-source and freely available at: