Distributed Deep Learning for IoT
N. Amuthan, N. Sivakumar, P. Velrajkumar · Advances in computational intelligence and robotics book series · 2022
Distributed deep learning is a type of machine learning that uses neural networks to learn and make predictions at scale. This is achieved by having many different computer systems that are connected via the internet. This allows for more parallel processing and faster results. In addition, when it comes to IoT, this type of technology can be used in conjunction with sensors and other devices to create more accurate predictions about the environment around us. Distributed deep learning can be used in many ways with the IoT because it can be applied to various aspects of IoT data processing, such as image recognition, speech recognition, natural language processing (NLP), or anomaly detection. The neural net is the most computationally intensive component of the system, and it requires a significant amount of energy. To make this system more cost-effective, there are two ways to lower the number of memory accesses: by reducing the size of images (so precision decreases), or by increasing network bandwidth so that there are fewer loop iterations required for each memory access.