A Hybrid Architecture for On-Device Compressive Machine Learning

Developing machine learning techniques that can protect the privacy of users’ data is of utmost importance, as tracking and selling our digital data (often without our permission and knowledge) has become a booming business model. By processing the data on the device that has collected the data, we can dramatically increase the level of privacy. On-device machine offers several additional benefits, such as low latency, efficient use of network bandwidth, and more autonomy. However, many devices deployed on the “edge” have very limited memory, weak processors, and scarce energy supply. This poses the challenge of envisioning new machine learning architectures that can function properly under such dire conditions. This issue is particularly urgent with the emergence of the Internet of Things. We propose a hybrid hardware-software framework that facilitates increased privacy protection due to on-device processing and moreover has the potential to significantly reduce the computational complexity and memory requirements of on-device machine learning. In the first step, inspired by compressive sensing, data is collected in compressed form simultaneously with the sensing process. Thus this compression happens already at the hardware level during data acquisition. But unlike in compressive sensing, this compression is achieved via a projection operator that is specifically tailored to the desired machine learning task. The second step consists of a specially designed and trained deep network. Numerical simulations in image classification illustrate the viability of our method.

PDF Abstract

Datasets


  Add Datasets introduced or used in this paper

Results from the Paper


  Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

Methods


No methods listed for this paper. Add relevant methods here