The edge platform has evolved to become a part of a distributed computing environment. While typical edges do not have enough processing power to train machine learning models in real time, it is common to generate models in the cloud for use on the edge. The pattern of heterogeneous Internet of Things (IoT) data is dependent on individual circumstances. It is not easy to guarantee prediction performance when a monolithic model is used without considering the spatial characteristics of the space generating those data. In this paper, we propose a collaborative framework using a new method to select the best model for the edge from candidate models of cloud based on sample data correlation. This method lets the edge use the most suitable model without any training tasks on the edge side, and it also minimizes privacy issues. We apply the proposed method to predict future fine particulate matter concentration in an individual space. The results suggest that our method can provide better performance than the previous method.

As deep learning technology paves its way, real-world applications that make use of it become popular these days. Edge computing architecture is one of the service architectures to realize the deep learning based service, which makes use of the resources near the data source or client. In Edge computing architecture it becomes important to manage resource usage, and there is research on optimization of deep learning, such as pruning or binarization, which makes deep learning models more lightweight, along with the research for the efficient distribution of workloads on cloud or edge resources. Those are to reduce the workload on edge resources. In this paper, a usage optimization method with batch and model management is proposed. The proposed method is to increase the utilization of GPU resource by modifying the batch size of the input of an inference application. To this end, the inference pipelines are identified to see how the different kinds of resources are used, and then the effect of batch inference on GPU is measured. The proposed method consists of a few modules, including a tool for batch size management which is able to change a batch size with respect to the available resources, and another one for model management which supports on-the-fly update of a model. The proposed methods are implemented on a real-time video analysis application and deployed in the Kubernetes cluster as a Docker container. The result shows that the proposed method can optimize the usage of edge resources for real-time video analysis deep learning applications.

With the proliferation of IoT devices, real-time data collection and monitoring services are being offered to various industries. However, time series data, due to its temporal nature, presents numerous unforeseen challenges over time, often leading to data contamination. Such contaminated data make it difficult to effectively utilize in practical applications. This paper investigates the issues with time series data collected from sensors and introduces preprocessing techniques and a corresponding platform to overcome these challenges. In addition to established preprocessing methods, novel approaches for processing and transforming multiple time series data sets are applied, followed by performance testing within the framework of the proposed service platform.