Programming by optimisation: Computer-aided design of high-performance algorithms for hard combinatorial problems

用于解决困难计算问题的高性能算法是现实应用中使用的许多复杂软件系统的核心。我研究的总体目标是开发和建立此类算法的开发范例，我称之为“优化编程（PBO）”。这种范式背后的关键思想是，人类专家指定的不是算法，而是潜在的大型算法设计空间，并使用自动化方法在该空间内找到性能优化的算法。这种方法使人类专家能够专注于思考解决给定问题的可能机制的创造性任务；同时，一旦指定了丰富且可能很大的设计空间，就可以针对特定类型的输入自动搜索其中的性能优化算法，从而可以用最少的人力获得针对不同应用情况的定制优化设计。

High-performance algorithms for solving hard computational problems are found at the core of many complex software systems used in real-world applications. The overall goal of my research is to develop and establish a paradigm for the development of such algorithms that I call "programming by optimisation (PBO)". The key idea behind this paradigm is for human experts to specify not an algorithm, but a potentially large design space of algorithms, and to use automated methods to find performance-optimised algorithms within this space. This approach allows human experts to focus on the creative task of thinking about possible mechanisms for solving given problems; at the same time, once a rich and potentially large design space has been specified, the automated search for a performance-optimised algorithm within it can be specific to a particular type of input, such that custom-optimised designs for different application situations can be obtained with minimal human effort.