Adaptive Approximate Accelerators with Controlled Quality Using Machine Learning

Approximate accelerators speed up regularly executed code. However, with an approximate static design, while the average output quality constraint is satisfied, the quality of individual outputs varies significantly with dynamically changing inputs. We propose to enhance the quality of approximation through design adaptation by predicting the most suitable settings of the approximate design to execute the inputs. The proposed method predicts the design settings based on the applied input data and user preferences, without losing the gains of approximations. We use machine learning (ML) algorithms to build an efficient and lightweight design selector to adapt the approximate accelerators to meet a user-defined quality constraint. We fully automate the proposed methodology of quality assurance of approximate accelerators using ML-based models, for both software and hardware implementations. The analysis results of image processing applications showed that it is possible to satisfy the target output quality with high accuracy.