Currently, Machine Learning (ML) is becoming ubiquitous in everyday life. Deep Learning (DL) is already present in many applications ranging from computer vision for medicine to autonomous driving of modern cars as well as other sectors in security, healthcare, and finance. However, to achieve impressive performance, these algorithms employ very deep networks, requiring a significant computational power, both during the training and inference time. A single inference of a DL model may require billions of multiply-and-accumulated operations, making the DL extremely compute-and energy-hungry. In a scenario where several sophisticated algorithms need to be executed with limited energy and low latency, the need for cost-effective hardware platforms capable of implementing energy-efficient DL execution arises. This paper first introduces the key properties of two brain-inspired models like Deep Neural Network (DNN), and Spiking Neural Network (SNN), and then analyzes techniques to produce efficient and high-performance designs. This work summarizes and compares the works for four leading platforms for the execution of algorithms such as CPU, GPU, FPGA and ASIC describing the main solutions of the state-of-the-art, giving much prominence to the last two solutions since they offer greater design flexibility and bear the potential of high energy-efficiency, especially for the inference process. In addition to hardware solutions, this paper discusses some of the important security issues that these DNN and SNN models may have during their execution, and offers a comprehensive section on benchmarking, explaining how to assess the quality of different networks and hardware systems designed for them.

Deep Neural Networks (DNNs) are nowadays a common practice in most of the Artificial Intelligence (AI) applications. Their ability to go beyond human precision has made these networks a milestone in the history of AI. However, while on the one hand they present cutting edge performance, on the other hand they require enormous computing power. For this reason, numerous optimization techniques at the hardware and software level, and specialized architectures, have been developed to process these models with high performance and power/energy efficiency without affecting their accuracy. In the past, multiple surveys have been reported to provide an overview of different architectures and optimization techniques for efficient execution of Deep Learning (DL) algorithms. This work aims at providing an up-to-date survey, especially covering the prominent works from the last 3 years of the hardware architectures research for DNNs. In this paper, the reader will first understand what a hardware accelerator is, and what are its main components, followed by the latest techniques in the field of dataflow, reconfigurability, variable bit-width, and sparsity.

In today’s world, ruled by a great amount of data and mobile devices, cloud-based systems are spreading all over. Such phenomenon increases the number of connected devices, broadcast bandwidth, and information exchange. These fine-grained interconnected systems, which enable the Internet connectivity for an extremely large number of facilities (far beyond the current number of devices) go by the name of Internet of Things (IoT). In this scenario, mobile devices have an operating time which is proportional to the battery capacity, the number of operations performed per cycle and the amount of exchanged data. Since the transmission of data to a central cloud represents a very energy-hungry operation, new computational paradigms have been implemented. The computation is not completely performed in the cloud, distributing the power load among the nodes of the system, and data are compressed to reduce the transmitted power requirements. In the edge-computing paradigm, part of the computational power is moved toward data collection sources, and, only after a first elaboration, collected data are sent to the central cloud server. Indeed, the “edge” term refers to the extremities of systems represented by IoT devices. This survey paper presents the hardware architectures of typical IoT devices and sums up many of the low power techniques which make them appealing for a large scale of applications. An overview of the newest research topics is discussed, besides a final example of a complete functioning system, embedding all the introduced features.

Transformers' compute- intensive operations pose enormous challenges for their deployment in resource- constrained EdgeAI / tiny ML devices. As an established neural network compression technique, quantization reduces the hardware computational and memory resources. In particular, fixed-point quantization is desirable to ease the computations using lightweight blocks, like adders and multipliers, of the underlying hardware. However, deploying fully-quantized Transformers on existing general-purpose hardware, generic AI accelerators, or specialized architectures for Transformers with floating-point units might be infeasible and/or inefficient. Towards this, we propose SwiftTron, an efficient specialized hardware accelerator designed for Quantized Transformers. SwiftTron supports the execution of different types of Transformers' operations (like Attention, Softmax, GELU, and Layer Normalization) and accounts for diverse scaling factors to perform correct computations. We synthesize the complete SwiftTron architecture in a 65 nm CMOS technology with the ASIC design flow. Our Accelerator executes the RoBERTa-base model in 1.83 ns, while consuming 33.64 mW power, and occupying an area of 273 mm ² • To ease the reproducibility, the RTL of our SwiftTron architecture is released at https://github.com/albertomarchisio/SwiftTron.

Falls in the home environment are a primary cause of injury in older adults. According to the U.S. Centers for Disease Control and Prevention, every year, one in four adults 65 years of age and older reports experiencing a fall. A variety of different technologies have been proposed to detect fall events. However, the need to detect all fall instances (i.e., to avoid false negatives) has led to the development of systems marked by high sensitivity and hence a significant number of false alarms. The occurrence of false alarms causes frequent and unnecessary calls to emergency response centers, which are critical resources that should be utilized only when necessary. Besides, false alarms decrease the level of confidence of end-users in the fall detection system with a negative impact on their compliance with using the system (e.g., wearing the sensor enabling the detection of fall events). Herein, we present a novel approach aimed to augment traditional fall detection systems that rely on wearable sensors and fall detection algorithms. The proposed approach utilizes a UWB-based tracking system and a home robot. When the fall detection system generates an alarm, the alarm is relayed to a base station that utilizes a UWB-based tracking system to identify where the older adult and the robot are so as to enable navigating the environment using the robot and reaching the older adult to check if he/she experienced a fall. This approach prevents unnecessary calls to emergency response centers while enabling a tele-presence using the robot when appropriate. In this paper, we report the results of a novel fall detection algorithm, the characteristics of the alarm notification system, and the accuracy of the UWB-based tracking system that we implemented. The fall detection algorithm displayed a sensitivity of 99.0% and a specificity of 97.8%. The alarm notification system relayed all simulated alarm notification instances with a maximum delay of 106 ms. The UWB-based tracking system was found to be suitable to locate radio tags both in line-of-sight and in no-line-of-sight conditions. This result was obtained by using a machine learning-based algorithm that we developed to detect and compensate for the multipath effect in no-line-of-sight conditions. When using this algorithm, the error affecting the estimated position of the radio tags was smaller than 0.2 m, which is satisfactory for the application at hand.

In the last years, the need for new efficient video compression methods grown rapidly as frame resolution has increased dramatically. The Joint Collaborative Team on Video Coding (JCT-VC) effort produced in 2013 the H.265/High Efficiency Video Coding (HEVC) standard, which represents the state of the art in video coding standards. Nevertheless, in the last years, new algorithms and techniques to improve coding efficiency have been proposed. One promising approach relies on embedding direction capabilities into the transform stage. Recently, the Steerable Discrete Cosine Transform (SDCT) has been proposed to exploit directional DCT using a basis having different orientation angles. The SDCT leads to a sparser representation, which translates to improved coding efficiency. Preliminary results show that the SDCT can be embedded into the HEVC standard, providing better compression ratios. This paper presents a hardware architecture for the SDCT, which is able to work at a frequency of 188 $\mathrm{M}$ $\mathrm{Hz}$ , reaching a throughput of 3.00 GSample/s. In particular, this architecture supports 8k UltraHigh Definition (UHD) (7680 × 4320) with a frame rate of 60 $\mathrm{Hz}$ , which is one of the best resolutions supported by HEVC.