Data compression for DNN weighting coefficients using layer adaptive quantization

2021 ◽  
Author(s):  
Ryota Aogaki ◽  
Yoshiyuki Yashima
Keyword(s):  
Data Compression ◽  
Adaptive Quantization ◽  
Weighting Coefficients

Fusion of dynamic predictive block adaptive quantization and vector quantization for staggered SAR data compression

2020 ◽  
Vol 12 (2) ◽  
pp. 206-215
Author(s):  
Hang Zou ◽  
Fengjun Zhao ◽  
Xiaoxue Jia ◽  
Heng Zhang ◽  
Wei Wang
Keyword(s):  
Data Compression ◽  
Vector Quantization ◽  
Adaptive Quantization ◽  
Sar Data

Study on an Algorithm for SAR Raw Data Compression

2013 ◽  
Vol 380-384 ◽  
pp. 1495-1498
Author(s):  
Shang Chun Zeng ◽  
Yun Xia Xie ◽  
Yi Xian Chen ◽  
Zhao Da Zhu
Keyword(s):  
Synthetic Aperture Radar ◽  
Data Compression ◽  
Linear Prediction ◽  
Synthetic Aperture ◽  
Bit Rate ◽  
Compression Method ◽  
Raw Data ◽  
Adaptive Quantization ◽  
Difference Series ◽  
Aperture Radar

t is difficult to directly compress the raw data of synthetic aperture radar for its low relativity. In this paper, a new algorithm is put forward. Firstly range focusing is imposed to SAR raw data, which makes it have comparative high relativity, secondly a linear prediction is performed along the azimuth, lastly block adaptive quantization is used to the prediction difference series. The experiments manifest that with same bit rate, SQNR and SDNR of the algorithm proposed in this paper surpass that of BAQ algorithm. The calculation in this paper is far less than that of compression method after range focusing advised in corresponding reference. The algorithm proposed in this paper has a certain practical value.

INFORMATION SYSTEMS OF ARTIFICIAL INTELLIGENCE MAKING THE BEST DECISIONS IN THE ECONOMY

2016 ◽  
Vol 4 (2) ◽  
pp. 197-203
Author(s):  
Александр Носов ◽  
Aleksandr Nosov
Keyword(s):  
Artificial Intelligence ◽  
Information Systems ◽  
Feature Space ◽  
Economic Systems ◽  
Practice Characteristic ◽  
Recognition Algorithms ◽  
Adaptive Quantization ◽  
Characteristic Values ◽  
Weighting Coefficients ◽  
Theoretical Foundations

The article examines the possibility of using artificial intelligence in economic systems. The theoretical foundations of the methodology description of recognizable objects in the selected feature space showed. An approved method of coding practice characteristic values based on their adaptive quantization, allowing learners to use simple recognition algorithms using the corrected weighting coefficients is given. Such systems can be used to address the macro- and micro-economic challenges as well as in shaping the competencies of staff.

Sign in / Sign up

Export Citation Format

Share Document