Integer Self-Organizing Maps for Digital Hardware

被引：0

作者：

Kleyko, Denis ^{[1
]}

Osipov, Evgeny ^{[1
]}

De Silva, Daswin ^{[2
]}

Wiklund, Urban ^{[3
]}

Alahakoon, Damminda ^{[2
]}

机构：

[1] Lulea Univ Technol, Lulea, Sweden

[2] La Trobe Univ, Melbourne, Vic, Australia

[3] Umea Univ, Umea, Sweden

来源：

2019 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN) | 2019年

基金：

瑞典研究理事会;

关键词：

Self-Organizing Maps; tri-state Self-Organizing Maps; FPGA; digital hardware; the clipping function; IP CORE; IMPLEMENTATION; ARCHITECTURE; PARALLEL; SOM;

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

The Self-Organizing Map algorithm has been proven and demonstrated to be a useful paradigm for unsupervised machine learning of two-dimensional projections of multidimensional data. The tri-state Self-Organizing Maps have been proposed as an accelerated resource-efficient alternative to the Self-Organizing Maps for implementation on field-programmable gate array (FPGA) hardware. This paper presents a generalization of the tri-state Self-Organizing Maps. The proposed generalization, which we call integer Self-Organizing Maps, requires only integer operations for weight updates. The presented experiments demonstrated that the integer Self-Organizing Maps achieve better accuracy in a classification task when compared to the original tri-state Self-Organizing Maps.

引用

页数：8

共 34 条

[1] Dynamic self-organizing maps with controlled growth for knowledge discovery [J].