Method to Improve UI Rendering using Predictive Sequence Modelling

Abstract

Google Play Store has over three million published applications, and each tries to provide distinctive User Interface (UI) layouts to enhance the user experience. A layout is a blueprint of the application UI comprising visual components of the interface. While building applications, developers strive to provide a richer user experience resulting in complicated and nested UI layouts at the cost of rendering performance. In this paper, we propose a system called Intelligent Resource Manager (IRM) composed of two components NxLPM (Next Layout Prediction Model) and ULP (UI Layout Pre-inflater), to improve UI layout rendering performance. IRM uses a data gatherer framework to collect and maintain a database with information on UI layouts and their parameters. NxLPM gets trained online based on the updated database and predicts the application's UI layouts with the help of sequence modelling. ULP ensures that the user interfaces corresponding to the predicted UI layouts are inflated earlier than required for either application launch or activity transition scenarios. IRM returns the already prepared UI layouts when the application demands, reducing the rendering time, thus improving the overall UI transition rates. Our experimental results showed an 18% gain in application launch time and a 32% reduction in activity transition inflate time with a prediction accuracy of 89%.