Complexity enabled design space exploration

摘要

The proposed paper presents a framework for complexity enabled design space exploration. A circuit problem is introduced where different designs are generated from a component library to create a low pass filter circuit. Performance and complexity metrics are defined to measure the fitness of the design. A scheme for complexity enabled design space exploration is introduced and the results obtained are compared to the traditional performance based exploration. It is found that the complexity enabled exploration results in identification of regions within the performance-complexity trade space which are difficult to obtain using a performance based exploration. Characteristic features of designs that result in high performance but low complexity are also identified. Later a scheme for local design space exploration is proposed which investigates the designs in the nearest neighborhood of a particular design. This provides insights about the flexibility and adaptability of the design.