Deep Neural Networks (DNNs) are very popular because of their high performance in various cognitive tasks in Machine Learning (ML). Recent advancements in DNNs have brought levels beyond human accuracy in many tasks, but at the cost of high computational complexity. To enable efficient execution of DNN inference, more and more research works, therefore, are exploiting the inherent error resilience of DNNs and employing Approximate Computing (AC) principles to address the elevated energy demands of DNN accelerators. This article provides a comprehensive survey and analysis of hardware approximation techniques for DNN accelerators. First, we analyze the state of the art, and by identifying approximation families, we cluster the respective works with respect to the approximation type. Next, we analyze the complexity of the performed evaluations (with respect to the dataset and DNN size) to assess the efficiency, potential, and limitations of approximate DNN accelerators. Moreover, a broad discussion is provided regarding error metrics that are more suitable for designing approximate units for DNN accelerators as well as accuracy recovery approaches that are tailored to DNN inference. Finally, we present how Approximate Computing for DNN accelerators can go beyond energy efficiency and address reliability and security issues as well.

/pdf/hardware-approximate-techniques-for-deep-neural-network-2f83pgtw.pdf

Hardware Approximate Techniques for Deep Neural Network Accelerators: A Survey

Deep neural networks (DNN) are now dominating in the most challenging applications of machine learning. As DNNs can have complex architectures with millions of trainable parameters (the so-called weights), their design and training are difficult even for highly qualified experts. In order to reduce human effort, neural architecture search (NAS) methods have been developed to automate the entire design process. The NAS methods typically combine searching in the space of candidate architectures and optimizing (learning) the weights using a gradient method. In this paper, we survey the key elements of NAS methods that – to various extents – consider hardware implementation of the resulting DNNs. We classified these methods into three major classes: single-objective NAS (no hardware is considered), hardware-aware NAS (DNN is optimized for a particular hardware platform), and NAS with hardware co-optimization (hardware is directly co-optimized with DNN as a part of NAS). Compared to previous surveys, we emphasize the multi-objective design approach that must be adopted in NAS and focus on co-design algorithms developed for concurrent optimization of DNN architectures and hardware platforms. As most research in this area deals with NAS for image classification using convolutional neural networks, we follow this trajectory in our paper. After reading the paper, the reader should understand why and how NAS and hardware co-optimization are currently used to build cutting-edge implementations of DNNs.

Neural Architecture Search and Hardware Accelerator Co-Search: A Survey

Recent technological (e.g., IoT, 5G) and economic (e.g., UN 2030 Sustainable Development Goals) developments have transformed the healthcare sector towards more personalized and IoT-based healthcare services. These services are realized through control and monitoring applications that are typically developed using artificial intelligence (AI)/machine learning (ML) based algorithms that play a significant role to highlight the efficiency of traditional healthcare systems. Current personalized healthcare services are dedicated in a specific environment and support technological personalization (e.g., personalized gadgets/devices) and are unable to consider different inter-related health conditions that lead to inappropriate diagnosis and affect sustainability and the long-term health/life of patients. Towards this problem, the state-of-the-art Healthcare 5.0 technology has evolved that supersede the previous healthcare technologies. The goal of healthcare 5.0 is to achieve a fully autonomous healthcare service, that takes into account the interdependent effect of different health conditions of a patient. This paper conducts a comprehensive survey on personalized healthcare services. In particular, we first present an overview of key requirements of comprehensive personalized healthcare services (CPHS) in modern healthcare Internet of Things (HIoT), including the definition of personalization and an example use case scenario as a representative for modern HIoT. Second, we explored a fundamental three-layer architecture for IoT-based healthcare systems using both AI/ML-based and non-AI-based approaches, considering key requirements for CPHS followed by their strengths and weaknesses in the frame of personalized healthcare services. Third, we highlighted different security threats against each layer of IoT architecture along with the possible AI-based and non-AI-based solutions. Finally, we proposed a methodology to provide reliable, resilient, and personalized healthcare services that address the identified weaknesses of existing approaches.

https://ieeexplore.ieee.org/ielx7/6287639/9668973/09658494.pdf

Reliable and Resilient AI and IoT-based Personalised Healthcare Services: A Survey

Recent technological (e.g., IoT, 5G), and economic (e.g., UN 2030 Sustainable Development Goals) developments have transformed the healthcare sector towards more personalized and IoT-based healthcare services. These services are realized through control and monitoring applications that are typically developed using artificial intelligence (AI)/machine learning (ML) based algorithms, that play a significant role to highlight the efficiency of traditional healthcare systems. Current personalized healthcare services are dedicated in a specific environment to support technological personalization (e.g., personalized gadgets/devices). However, they are unable to consider different inter-related health conditions, leading to inappropriate diagnosis and affect sustainability and the long-term health/life of patients. Towards this problem, the state-of-the-art Healthcare 5.0 technology has evolved that supersede previous healthcare technologies. The goal of healthcare 5.0 is to achieve a fully autonomous healthcare service, that takes into account the interdependent effect of different health conditions of a patient. This paper conducts a comprehensive survey on personalized healthcare services. In particular, we first present an overview of key requirements of comprehensive personalized healthcare services (CPHS) in modern healthcare Internet of Things (HIoT), including the definition of personalization and an example use case scenario as a representative for modern HIoT. Second, we explored a fundamental three-layer architecture for IoT-based healthcare systems using both AI and non-AI-based approaches, considering key requirements for CPHS followed by their strengths and weaknesses in the frame of personalized healthcare services. Third, we highlighted different security threats against each layer of IoT architecture along with the possible AI and non-AI-based solutions. Finally, we propose a methodology to develop reliable, resilient, and personalized healthcare services that address the identified weaknesses of existing approaches. 

/pdf/reliable-and-resilient-ai-and-iot-based-personalised-3v1jfs9j.pdf

Reliable and Resilient AI and IoT-Based Personalised Healthcare Services: A Survey

Computer-aided design (CAD) is the use of computer-based software to aid in design modeling, design analysis, design review, and design documentation. Nevertheless, the benefits of CAD can be elevated in combination with artificial intelligence (AI), extended reality, and manufacturing. AI can create an intelligent graphics interface and change tedious design processes into sophisticated ones. In extended reality technology, simulation can take place in a 3D virtual environment, thereby providing excellent interaction and better analysis. In manufacturing, as seen in 3D printing technology, CAD systems can be directly connected to manufacturing to produce complex parts easily and rapidly. In this paper, the integration of artificial intelligence (AI) in CAD, as well as CAD application in extended reality and 3D printing is examined. The primary aim of this review is to present an overview of current state-of-the-art CAD and its applications, as well as to forecast its future prospects. The article is written using a systematic review of journal papers with a focus on a wide spectrum of potentially relevant researches on CAD. The benefits of incorporating AI into the CAD systems, the use of CAD in extended reality and 3D printing, and finally a brief discussion of issues that are pushing CAD to new levels are all discussed. Finally, the review concluded that the demand for several varied products based on a single object input, immersive and interactive simulation, and direct design-to-manufacturing integration is driving CAD to new levels. • Incorporating AI into the CAD system will reduce the lead time dramatically and creates knowledge-based design environment. • Utilizing CAD for 3D printing allows design flexibility, enabling the fabrication of complex items that would be impossible to construct with traditional machining methods. • Through immersive simulation, VR, AR, and MR provide detailed conceptualization of various complicated pro-cuts that exist in the real world. 

Future prospects of computer-aided design (CAD) – A review from the perspective of artificial intelligence (AI), extended reality, and 3D printing

Advances in deep neural networks (DNNs) and the availability of massive real-world data have enabled superhuman levels of accuracy on many AI tasks and ushered the explosive growth of AI workloads across the spectrum of computing devices. However, their superior accuracy comes at a high computational cost, which necessitates approaches beyond traditional computing paradigms to improve their operational efficiency. Leveraging the application-level insight of error resilience, we demonstrate how approximate computing (AxC) can significantly boost the efficiency of AI platforms and play a pivotal role in the broader adoption of AI-based applications and services. To this end, we present RaPiD, a multi-tera operations per second (TOPS) AI hardware accelerator core (fabricated at 14-nm technology) that we built from the ground-up using AxC techniques across the stack including algorithms, architecture, programmability, and hardware. We highlight the workload-guided systematic explorations of AxC techniques for AI, including custom number representations, quantization/pruning methodologies, mixed-precision architecture design, instruction sets, and compiler technologies with quality programmability, employed in the RaPiD accelerator.

Ankur Agarwal

Papers

Efficient AI System Design With Cross-Layer Approximate Computing