What is it about?
This paper introduces three effective algorithms to check and improve data quality by fixing or imputing incorrect and missing data. We thoroughly tested these algorithms on TinyML devices and used real-world data to simulate data streams in real-time. The algorithms performed well in various tests, showing they are energy-efficient, fast, and accurate. Interestingly, our findings showed that even on less expensive hardware, these algorithms matched the performance of more advanced methods used on expensive devices. Additionally, they can clean data directly on the devices, reducing reliance on cloud computing and enabling immediate data processing and quality checks.
Featured Image
Photo by Hamed Taha on Unsplash
Why is it important?
first work to introduce imputation directly on TinyML boards and target real-time deployment and direct data processwing at IoT level.
Perspectives
it was very challenging but enjoyable, working with TinyML boards was not easy both from constraints and programming point of views, but the end result was really satisfying.
Shamil Al-Ameen
University of Mosul
Read the Original
This page is a summary of: Tiny-Impute: A Framework for On-device Data Quality Validation, Hybrid Anomaly Detection, and Data Imputation at the Edge, December 2023, ACM (Association for Computing Machinery),
DOI: 10.1145/3603166.3632164.
You can read the full text:
Contributors
The following have contributed to this page
