Poster: Approximate Memoization for Perception-based Mobile Applications

One of the main thrusts of mobile and pervasive computing is supporting perception-based applications [1]. Perceptionbased applications are those that help users augment their understanding of the physical world through the sensors on their mobile devices, e.g. augmented reality, visual product search, speech-to-text. Although mobile devices now have multi-core CPUs and multi-GB RAMs, these applications cannot be executed entirely on the devices. These applications need intensive computation and access to “big data” for them to be fast and accurate. They rely on offloading intensive tasks to the cloud. The devices send sensed values to the cloud, which then executes the recognition procedures using its computational resources [1] and access to big data. However, the heavy computation and the added communication latency still deter seamless interaction, which is desired for such applications. Hence, there is a need to accelerate the performance of perception-based mobile applications. In this regard, we believe approximate memoization will be a key enabling-technique.