SSL: Synchronous Self-paced Learning for Internet-of-Things Devices

With the emergence of Internet-of-Things (IoT), we are witnessing rapid increases in the amount of IoT devices. However, due to IoT devices are often deployed in dynamic and unmanned environment, it is imperative that the IoT devices have automatic model construction capabilities. In this paper, we propose a novel two-step approach called synchronous self-paced learning (SSL) for the IoT devices to construct the model automatically. In the first step, we design a synchronous mechanism that makes correlated devices synchronized and obtain the pseudo labels to form the original training set. In the second step, we propose a sample selection method based on reliability and diversity to filter the original training set, based on which we automatically accomplish the model construction. We conduct empirical evaluation of our SSL method compared with three state-of-the-art methods; and the evaluation shows that by applying our SSL method, we can achieve the classifier with higher accuracy.