Challenges to implement Machine Learning in Embedded Systems

Machine Learning has an important role in extraction of useful information out of sextillion bytes of data collected per day. The focus of machine learning is the development of computer programs that can access data and use it to learn for themselves. In some applications the goal of machine learning is to identify patterns by analyzing them (for e.g. in image recognition, surveillance); while in others the aim is to take actions based upon input data (for e.g. in robotics, self-driving cars). For many of such applications embedded systems are used over cloud due to latency, privacy, and security concerns. Also using machine learning with embedded system creates a problem of restricted memory and processor speed. Although the required memory and processing speed can be met with high performance hardware but still time, space, cost, security, privacy and power consumption create a challenge in implementing machine learning in embedded system. In this paper it is being discussed about these challenges and how these challenges can be addressed using various techniques.