Search for: All records

We present POPPER, a dataflow system for building Machine Learning (ML) workflows. A novel aspect of POPPER is its built-in support for in-flight error handling, which is crucial in developing effective ML workflows. POPPER provides a convenient API that allows users to create and execute complex workflows comprising traditional data processing operations (such as map, filter, and join) and user-defined error handlers. The latter enables inflight detection and correction of errors introduced by ML models in the workflows. Inside POPPER, we model the workflow as a reactive dataflow, a directed cyclic graph, to achieve efficient execution through pipeline parallelization. We demonstrate the in-flight error-handling capabilities of POPPER, for which we have built a graphical interface, allowing users to specify workflows, visualize and interact with its reactive dataflow, and delve into the internals of POPPER. 

Mobile app energy profilers provide a foundational energy diagnostic tool by identifying energy hotspots in the app source code. However, they only tackle the first challenge faced by developers, as, after presented with the energy hotspots, developers typically do not have any guidance on how to proceed with the remaining optimization process: (1) Is there a more energy-efficient implementation for the same app task? (2) How to come up with the more efficient implementation? To help developers tackle these challenges, we developed a new energy profiling methodology called differential energy profiling that automatically uncovers more efficient implementations of common app tasks by leveraging existing implementations of similar apps which are bountiful in the app marketplace. To demonstrate its effectiveness, we implemented such a differential energy profiler, DIFFPROF, for Android apps and used it to profile 8 groups (from 6 popular app categories) of 5 similar apps each. Our extensive case studies show that DIFFPROF provides developers with actionable diagnosis beyond a traditional energy profiler: it identifies non-essential (unmatched or extra) and known-to-be inefficient (matched) tasks, and the call trees of tasks it extracts further allow developers to quickly understand the reasons and develop fixes for the energy difference with minor manual debugging efforts.