On-device training of machine learning models on microcontrollers with federated learning

N Llisterri Giménez, M Monfort Grau… - Electronics, 2022 - mdpi.com
Recent progress in machine learning frameworks has made it possible to now perform
inference with models using cheap, tiny microcontrollers. Training of machine learning …

Automated log-scale quantization for low-cost deep neural networks

S Oh, H Sim, S Lee, J Lee - … of the IEEE/CVF Conference on …, 2021 - openaccess.thecvf.com
Quantization plays an important role in deep neural network (DNN) hardware. In particular,
logarithmic quantization has multiple advantages for DNN hardware implementations, and …

A hardware-friendly low-bit power-of-two quantization method for cnns and its fpga implementation

X Sui, Q Lv, Y Bai, B Zhu, L Zhi, Y Yang, Z Tan - Sensors, 2022 - mdpi.com
To address the problems of convolutional neural networks (CNNs) consuming more
hardware resources (such as DSPs and RAMs on FPGAs) and their accuracy, efficiency …

Image-captioning model compression

V Atliha, D Šešok - Applied Sciences, 2022 - mdpi.com
Image captioning is a very important task, which is on the edge between natural language
processing (NLP) and computer vision (CV). The current quality of the captioning models …

Hardware-centric automl for mixed-precision quantization

K Wang, Z Liu, Y Lin, J Lin, S Han - International Journal of Computer …, 2020 - Springer
Abstract Model quantization is a widely used technique to compress and accelerate deep
neural network (DNN) inference. Emergent DNN hardware accelerators begin to support …

Integer-only cnns with 4 bit weights and bit-shift quantization scales at full-precision accuracy

M Vandersteegen, K Van Beeck, T Goedemé - Electronics, 2021 - mdpi.com
Quantization of neural networks has been one of the most popular techniques to compress
models for embedded (IoT) hardware platforms with highly constrained latency, storage …

Neuron-by-Neuron Quantization for Efficient Low-Bit QNN Training

A Sher, A Trusov, E Limonova, D Nikolaev… - Mathematics, 2023 - mdpi.com
Quantized neural networks (QNNs) are widely used to achieve computationally efficient
solutions to recognition problems. Overall, eight-bit QNNs have almost the same accuracy …

Ax-BxP: Approximate blocked computation for precision-reconfigurable deep neural network acceleration

R Elangovan, S Jain, A Raghunathan - ACM Transactions on Design …, 2022 - dl.acm.org
Precision scaling has emerged as a popular technique to optimize the compute and storage
requirements of Deep Neural Networks (DNNs). Efforts toward creating ultra-low-precision …

Adaptive Global Power-of-Two Ternary Quantization Algorithm Based on Unfixed Boundary Thresholds

X Sui, Q Lv, C Ke, M Li, M Zhuang, H Yu, Z Tan - Sensors, 2023 - mdpi.com
In the field of edge computing, quantizing convolutional neural networks (CNNs) using
extremely low bit widths can significantly alleviate the associated storage and computational …

Quantized Graph Neural Networks for Image Classification

X Xu, L Ma, T Zeng, Q Huang - Mathematics, 2023 - mdpi.com
Researchers have resorted to model quantization to compress and accelerate graph neural
networks (GNNs). Nevertheless, several challenges remain:(1) quantization functions …