**Preface to "Recent Advances in Embedded Computing, Intelligence and Applications"**

The latest proliferation of Internet of Things deployments and edge computing combined with artificial intelligence has led to new exciting application scenarios, where embedded digital devices are essential enablers. Moreover, new powerful and efficient devices are appearing to cope with workloads formerly reserved for the cloud, such as deep learning. These devices allow processing close to where data are generated, avoiding bottlenecks due to communication limitations. The efficient integration of hardware, software and artificial intelligence capabilities deployed in real sensing contexts empowers the edge intelligence paradigm, which will ultimately contribute to the fostering of the offloading processing functionalities to the edge.

In this book, researchers have contributed nine peer-reviewed papers covering a wide range of topics in the area of edge intelligence. Among them are hardware-accelerated implementations of deep neural networks, IoT platforms for extreme edge computing, neuro-evolvable and neuromorphic machine learning, and embedded recommender systems. In chapters 1 and 2, the importance of speeding up the design space exploration and the optimization of reconfigurable accelerations for neural networks is introduced to the readers and addressed by proposing a novel method for improving the performance estimation of key metrics during the design space exploration, as well as the evaluation of multi-objective evolutionary algorithms with quantization in real hardware acceleration platforms.

Moreover, in chapters 3, 4, and 5, the topic of low power design for the Fog, Edge and Extreme Edge layers is presented to the readers from three main perspectives: low-latency wireless communication considering energy management strategies on hardware platforms; the modularity and flexibility of the sensor nodes for resource-constrained distributed computing with enhanced security; and energy-aware machine learning strategies through neuromorphic hardware. Chapter 6 reinforces the idea of scalable neuroevolution in hardware by proposing a dynamically reconfigurable block-based neural network model integrated with an evolutionary algorithm implemented in hardware.

Finally, chapters 7, 8, and 9 show three exciting and diverse investigations where embedded systems and embedded intelligence techniques are applied to recommender systems, hyperspectral image processing for brain cancer classification, and 2D graphic accelerators embedded in hardware platforms. This way, the benefits of using hardware acceleration techniques close to the data sources are presented to the readers.

The guest editors would like to thank all the authors who contributed to the Special Issue for their very high-quality research works in embedded computing, intelligence, and applications. The underlying advances and actual implementations allow significant progress to be made in the related application domains.

> **Jorge Portilla, Andres Otero, and Gabriel Mujica** *Editors*
