GADGET: A Toolkit for Optimization-Based Approaches to Interface and Display Generation

摘要

Recent work is beginning to reveal the potential of numerical optimization as an approach to generating interfaces and displays. Optimization-based approaches can often allow a mix of independent goals and constraints to be blended in ways that are difficult to describe algorithmically. While optimization-based techniques appear to offer several potential advantages, further research in this area is hampered by the lack of appropriate tools. Optimization toolkits do exist, but they typically require substantial specialized knowledge because they have been designed for traditional optimization problems. GADGET is an experimental toolkit to support optimization as an approach to interface and display generation. GADGET provides three core abstractions, initializers, iterations, and evaluations. An initializer creates an initial solution to be optimized, based on an existing algorithm or randomly. Iterations are responsible for transforming one potential solution into another, typically using methods that are at least partially random. Finally, evaluations are used for judging the different notions of goodness in a solution. Together with a evaluation standardization framework, support for generic properties integrated with an efficient lazy evaluation framework, and a library of reusable iterations and evaluations, the abstractions provided by GADGET simplify the development of optimization-based approaches to interface and display generation.