LunchBoxML exposes new components for training and using several general‐purpose Machine Learning algorithms in the context of computational design software. LunchBox relies on a C# machine learning framework called Accord.NET which is a “framework for building production‐grade computer vision, computer audition, signal processing, and statistics applications even for commercial use.” Accord.NET provides extensive documentation and sample applications for creating custom implementations of this framework.
LunchBoxML exposes several machine learning algorithms using Accord.NET.
Components, such as the Naïve Bayes Classification component, are examples of a supervised learning algorithm which requires users to establish “labeled” training datasets that the solver then uses to make predicted classifications. The K‐means and Gaussian Mixture components are examples of unsupervised learning methods that don’t require pre‐labeled datasets to derive groupings based on implicit characteristics in the data.
Given the variety of machine learning components available, the selection of the appropriate machine learning algorithm is based on the problem that the designer is hoping to solve. The effectiveness of the algorithm’s predictive capability is based on the quality of the inputs provided by the designer.