Deep Neural Network (DNN) inference on edge devices is now a common practice. However, tailoring a model for multiple devices involves a lot of time and effort. While elastic models, also known as weight-sharing models, have been proposed as an efficient solution to create high-accuracy, low-latency modules, the design of an elastic model's candidate space (search space) has been underexplored. We identified a new characteristic in candidate spaces, which we named sensitivity, made a design rationale for candidate spaces based on it, and then built a preliminary algorithm to generate candidate spaces. Results show that we can get a range of models (with a 2.75× FLOPs range) on the Pareto frontier of the space by training only once.

The automotive navigation often embarrasses drivers by providing guidance that is awkward, incomprehensible, or almost impossible to follow. We point out the lack of on-the-spot awareness as the key reason behind this situation. The current navigation does not consider a driver's characteristics such as driving ability, as well as the detailed conditions of the current driving environment. In this paper, we explore the cases of embarrassment related to navigation usage. We collected a total of 56 cases of embarrassing moments from three drivers' experiences and derived 8 categories of embarrassing moments.