Data Compression and Damage Evaluation of Underground Pipeline With Musicalized Sonar GMM

Tao, Kai; Wang, Qiang; Yue, Dong

doi:10.1109/tie.2023.3270519

Cited by 6 publications

(1 citation statement)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Consequently, these issues hinder energy efficiency and latency-effectiveness in the implementation of DNNs. To address these challenges, researchers have dedicated various efforts to minimize the main memory footprint of DNNs through approaches such as reducing model size or optimizing memory access [8], [45], [47].…”

Section: Introductionmentioning

confidence: 99%

O-2A: Outlier-Aware Compression for 8-bit Post-Training Quantization Model

Ho,

Chang

2023

IEEE Access

View full text Add to dashboard Cite

Post Training Quantization (PTQ) is a practical and cost-effective technique that reduces main memory footprint of Deep Neural Networks (DNNs). However, the effectiveness of PTQ is limited by a notable decrease in accuracy when the precision falls below 8 bits. To overcome this limitation of PTQ, we present a new compression method called Outlier Aware Approximation (O-2A) that compresses 8-bit PTQ models into lower precision with minimal accuracy loss. In O-2A, parameters are classified into outliers and non-outliers. Critical bits of outliers are preserved, while unnecessary bits of non-outliers are removed, resulting in reduced compression error. Nevertheless, compressing outliers using O-2A below 6 bits precision poses challenges due to significant compression error. To achieve more aggressive compression, we introduce multiple levels of O-2A (mO-2A), where outliers are divided into different levels, and compression error is minimized. We evaluate our techniques on the ImageNet dataset using the Pytorch framework. The results demonstrate that our methods outperform previous works in term of accuracy with the same compression efficiency.

show abstract

Section: Introductionmentioning

confidence: 99%

O-2A: Outlier-Aware Compression for 8-bit Post-Training Quantization Model

Ho,

Chang

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

Underground concrete pipe crack damage monitoring by the fuzzy analysis of microseism

Tao,

Jiang

2024

Meas. Sci. Technol.

View full text Add to dashboard Cite

Underground concrete pipes are critical for the safety of urban infrastructure. Cracks could lead to accidents such as fluid spillage and ground subsidence, as well as waterlogging. Thus, it is of great significance to measure and monitor the crack of pipeline. In this study, a fuzzy monitoring method for concrete pipe crack damage based on the microseism signal Gaussian Mixed Model (GMM) analysis was proposed. First, the multiple parameters of the microseism signal were extracted. Then, Criteria Importance Through Intercriteria Correlation (CRITIC) weight vector was constructed using the importance information of parameters. Furthermore, the GMM membership matrix was created by the similarity of the probability density distribution. The crack condition was assessed by means of a fuzzy calculation between the weight vector and membership matrix. The experiment indicated that the proposed method can monitor and assess the different crack conditions in real time.

show abstract

Energy Consumption Optimization for Heating, Ventilation and Air Conditioning Systems Based on Deep Reinforcement Learning

Shen

Tang

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Data Compression and Damage Evaluation of Underground Pipeline With Musicalized Sonar GMM

Cited by 6 publications

References 34 publications

O-2A: Outlier-Aware Compression for 8-bit Post-Training Quantization Model

O-2A: Outlier-Aware Compression for 8-bit Post-Training Quantization Model

Underground concrete pipe crack damage monitoring by the fuzzy analysis of microseism

Energy Consumption Optimization for Heating, Ventilation and Air Conditioning Systems Based on Deep Reinforcement Learning

Contact Info

Product

Resources

About