Resilience Path: Paper to WWW 2016 Accepted
Published on in RESILIENCE PATH
After the success of INFOCOM’16, the Resilience Path of cfaed celebrates yet another top-tier publication: A paper on the integration of incremental and approximate computations has been accepted to WWW 2016, a leading conference in the area of "Big Data" analytics. This paper is especially important for our efforts to strengthen the 5G Lab, and the HAEC initiatives at cfaed to support the development of the Tactile Internet. In particular, the proposed data analytics system, called IncApprox, uses a combination of incremental and approximate computing paradigms to enable low-latency energy efficient stream processing.
The paper presents IncApprox, a data analytics system that combines the benefits of incremental and approximate computing paradigms. The paper builds on the observation that both paradigms rely on computing over a subset of data items instead of computing over the entire dataset, but they differ in their means for skipping parts of the computation. To marry these two complementary paradigms, the paper presents an online stratified sampling algorithm that uses self-adjusting computation to produce an incrementally updated approximate output with bounded error.