Distributed Communicating Neural Network Architecture for Smart Environments

Abstract

The deployment of millions of embedded sensors plagued by resource constraints in sophisticated, complex and dynamic IoT smart environments continues to inspire the need to build novel architectures and models for automated, efficient inference and communication in distributed smart settings. In such settings, practical challenges related to energy efficiency, computational power and reliability, tedious design implementation, effective communication, optimal sampling and accurate event classification, prediction and detection exist. Sensors operating in smart environments must be capable of overcoming such challenges and enable scalable monitoring of dynamic phenomena while conducting real-time operations. The development of Machine Learning (ML) continues to motivate a new wave of innovative solutions that intermarry embedded sensors, IoT, and ML to enable various applications in smart environments. We propose a distributed communicating architecture based on Recurrent Neural Networks (RNNs) that can be instantiated on smart devices observing unique data and performing automated distributed inference via hidden-state communication. Our model uses a data-driven approach to collectively solve various distributed objectives, as evidenced by a series of systematic analyses we present. Although demonstrated on a small setup (2/3) nodes, this work sets out a new direction for automatically learning to communicate to solve tasks in distributed settings.