RNSiM: Efficient Deep Neural Network Accelerator Using Residue Number Systems

Roohi, Arman; Taheri, MohammadReza; Angizi, Shaahin; Fan, Deliang

doi:10.1109/iccad51958.2021.9643531

Cited by 15 publications

(4 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Commonly used nonlinear activation functions include softmax, logistic, hyperbolic, and ReLU [22]. In the context of neural network development with the RNS system, advantages have been observed in performing addition, subtraction, and multiplication operations [15].…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Support vector machines implementation over integers modulo-M and Residue Number System

Arenas-Hoyos,

Bernal-Noreña

2023

DYNA

View full text Add to dashboard Cite

In low-power hardware implementations for classification algorithms, it is often essential to use physical resources efficiently. In this sense, the use of modulo-M integer operations instead of floating-point arithmetic, can lead to better performance, especially when M represents the dynamic range of an arithmetic block of the Residue Number System (RNS) [1,2]. Following this premise, this work is aiming to provide a methodology for implementing a classifier, specifically a Support Vector Machine (SVM) [3], using modulo-M integers and proposing a method for the use of Residue Number System.

show abstract

Section: Discussionmentioning

confidence: 99%

“…Additionally, the research presents a methodology for performing the exponential operation in RNS. Both of these objectives hold fundamental importance in advancing the practical application of unconventional numerical systems, such as RNS, in pattern recognition systems and digital signal processing [11,15,16,24,26].…”

Section: Discussionmentioning

confidence: 99%

Support vector machines implementation over integers modulo-M and Residue Number System

Arenas-Hoyos,

Bernal-Noreña

2023

DYNA

View full text Add to dashboard Cite

show abstract

“…A complete multiplier-free RNS accelerator utilizing this approach is also developed [14]. RNS has been also utilized in the design of in-memory computing (IMC) systems [15], [16]. This work addresses the shortcomings of existing state-of-the-art RNS DNN architectures (Sec.…”

Section: B Rns In Dnn Acceleratorsmentioning

confidence: 99%

A 22-nm 4.92 TOPS/W end-to-end RNS DNN Accelerator for Edge-AI Devices

Sakellariou,

Paliouras,

Kouretas

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…The authors observed that the quantized model produced better accuracy results than the fullprecision model. Quantization on CNNs applied to portable devices is discussed in [22][23][24][25], where the authors evaluated inference results on medical images using a quantized CNN and exhibited that the inference time is reduced by 97%. The objective of quantizing a neural network is to decrease the bit-width of the network parameters: weights and/or activation.…”

Section: Hardware-friendly Neural Networkmentioning

confidence: 99%

Enabling Intelligent IoTs for Histopathology Image Analysis Using Convolutional Neural Networks

et al. 2022

Self Cite

View full text Add to dashboard Cite

Medical imaging is an essential data source that has been leveraged worldwide in healthcare systems. In pathology, histopathology images are used for cancer diagnosis, whereas these images are very complex and their analyses by pathologists require large amounts of time and effort. On the other hand, although convolutional neural networks (CNNs) have produced near-human results in image processing tasks, their processing time is becoming longer and they need higher computational power. In this paper, we implement a quantized ResNet model on two histopathology image datasets to optimize the inference power consumption. We analyze classification accuracy, energy estimation, and hardware utilization metrics to evaluate our method. First, the original RGB-colored images are utilized for the training phase, and then compression methods such as channel reduction and sparsity are applied. Our results show an accuracy increase of 6% from RGB on 32-bit (baseline) to the optimized representation of sparsity on RGB with a lower bit-width, i.e., <8:8>. For energy estimation on the used CNN model, we found that the energy used in RGB color mode with 32-bit is considerably higher than the other lower bit-width and compressed color modes. Moreover, we show that lower bit-width implementations yield higher resource utilization and a lower memory bottleneck ratio. This work is suitable for inference on energy-limited devices, which are increasingly being used in the Internet of Things (IoT) systems that facilitate healthcare systems.

show abstract

RNSiM: Efficient Deep Neural Network Accelerator Using Residue Number Systems

Cited by 15 publications

References 19 publications

Support vector machines implementation over integers modulo-M and Residue Number System

Support vector machines implementation over integers modulo-M and Residue Number System

A 22-nm 4.92 TOPS/W end-to-end RNS DNN Accelerator for Edge-AI Devices

Enabling Intelligent IoTs for Histopathology Image Analysis Using Convolutional Neural Networks

Contact Info

Product

Resources

About