Deep Learning in IoT: Introduction, Applications, and Perspective in the Big Data Era

Introduction

The era of the IoT and big data is creating and acquiring an interest in diverse research disciplines, which can be seen in the amount of recently published articles, surveys, and tutorials on the topic (Chen, 2014; Gheisari, 2017; Horidi, 2017; Najafabadi, 2015). According to McKinsey’s report on the global economic impact of the IoT (Najafabadi, 2015), the annual economic impact of the IoT in 2025 will be in the range of $2.7 to $6.2 trillion. Healthcare applications contribute about 41% of the share in the IoT market, followed by industry and energy with 33% and 7%, respectively. Other fields make up about 15% of the IoT market. Deep learning (DL) implies a tremendous and sudden growth in the market in the coming years (Horidi, 2017) since DL is suitable for most of the real-time applications. This has led to increased demand for DL products and tasks, and experts are looking forward to designing these new products. McKinsey’s report also mentions that advances in DL algorithms are the main enablers of knowledge work automation. Over a decade, IoT applications have become spread across many fields (health, transportation, smart home, smart city, agriculture, education, and others) that consider key design elements as intelligent learning behaviors such as prediction, mining data, and pattern recognition. DL, which is a part of the machine learning (ML) approach, is actively utilized in IoT applications compared to other methods. DL is getting prominence by its intrinsic behavior of analytics, as it is easier to integrate big data when compared to traditional ML. Based on the learning and hierarchical representation of deep architecture, DL consists of supervised and unsupervised learning techniques for the design structure of neural networks. The functioning of DL imitates human brain neurons in the transformation of data. The IoT and big data are inter-related. IoT directly or indirectly generates big data and the IoT is an important target for big data analytics, which mainly aims to improve IoT services (Goodfello, 2016). Searching trends in Google demonstrates the popularity of DL compared to the other four ML techniques. The definitions of these recent technologies are given by different organizations and are summarized in Table 1.