Efficient Hardware Architectures for Accelerating Deep Neural Networks: Survey

Abstract

In the modern-day era of technology, a paradigm shift has been witnessed in the areas involving applications of Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL). Specifically, Deep Neural Networks (DNNs) have emerged as a popular field of interest in most AI applications such as computer vision, image and video processing, robotics, etc. In the context of developed digital technologies and the availability of authentic data and data handling infrastructure, DNNs have been a credible choice for solving more complex real-life problems. The performance and accuracy of a DNN is a way better than human intelligence in certain situations. However, it is noteworthy that the DNN is computationally too cumbersome in terms of the resources and time to handle these computations. Furthermore, general-purpose architectures like CPUs have issues in handling such computationally intensive algorithms. Therefore, a lot of interest and efforts have been invested by the research fraternity in specialized hardware architectures such as Graphics Processing Unit (GPU), Field Programmable Gate Array (FPGA), Application Specific Integrated Circuit (ASIC), and Coarse Grained Reconfigurable Array (CGRA) in the context of effective implementation of computationally intensive algorithms. This paper brings forward the various research works on the development and deployment of DNNs using the aforementioned specialized hardware architectures and embedded AI accelerators. The review discusses the detailed description of the specialized hardware-based accelerators used in the training and/or inference of DNN. A comparative study based on factors like power, area, and throughput, is also made on the various accelerators discussed. Finally, future research and development directions, such as future trends in DNN implementation on specialized hardware accelerators, are discussed. This review article is intended to guide hardware architects to accelerate and improve the effectiveness of deep learning research.INDEX TERMS Machine learning, field programmable gate array (FPGA), deep neural networks (DNN), deep learning (DL), application specific integrated circuits (ASIC), artificial intelligence (AI), central processing unit (CPU), graphics processing unit (GPU), hardware accelerators. FIGURE 3. A single ANN neuron with its elements (inputs, weights, bias, summer, activation function, and output).

QoS-Aware Inference Acceleration Using Adaptive Depth Neural Networks

“…However, the policies learned for the decision networks are opaque to users, and, hence, they cannot be used for predictable adaptation to meet the resource conditions. Various hardware-and software-based inference accelerators have been developed to run trained neural networks efficiently on target devices [43], [44]. While hardware-based accelerators try to maximize the throughput of deep learning operations on specialized hardware [45], software-based accelerators mainly focus on optimizing resource management, pipeline design, model restructuring, and quantization [46]- [50].…”

Accelerated Low Power AI System for Indian Sign Language Recognition

“…FPGA-based DNN accelerators [13], can be broadly categorized into two types: accelerators tailored for specific applications such as speech recognition, object detection, and natural language processing, and accelerators designed for specific algorithms such as CNN and RNN. Additionally, there exist accelerator frameworks equipped with hardware templates.…”

Deep learning for economic transformation: a parametric review

“…Innovations in automated machine learning (AutoML) [25] and continual learning models, model compression for deployment on resource-limited devices, robustness against data distribution shifts, and effective multi-modal data integration are pivotal for maximizing DL's impact. For that reason, optimal DL model performance necessitates co-designing hardware [26] and software, highlighting the intricate balance between technological advancements and practical applications in economic contexts. All the preceding and succeeding data is context based on following envisioned environment, as illustrated in Table 1.…”