Related Experiment Video
Updated: Mar 31, 2026
08:59
Morphology-Based Distinction Between Healthy and Pathological Cells Utilizing Fourier Transforms and Self-Organizing Maps
Published on: October 28, 2018
7.7K
Improved Learning Performance of Hardware Self-Organizing Map Using a Novel Neighborhood Function.
IEEE Transactions on Neural Networks and Learning Systems
|October 21, 2015
Summary
This study introduces a new hardware-friendly neighborhood function for self-organizing maps (SOMs) to enhance vector quantization. The novel function improves SOM performance without increasing hardware costs or reducing speed.
Area of Science:
- Artificial Intelligence
- Machine Learning
- Hardware Acceleration
Background:
- Traditional hardware implementations of self-organizing maps (SOMs) often use neighborhood functions limited to negative powers of two.
- This limitation can impact the vector quantization performance of hardware-based SOMs.
Purpose of the Study:
- To propose and evaluate a novel, hardware-friendly neighborhood function for SOMs.
- The goal is to improve vector quantization performance in hardware SOM implementations.
Main Methods:
- The proposed neighborhood function was simulated to assess its vector quantization capabilities.
- The hardware SOM with the new function was implemented on a field-programmable gate array (FPGA) to evaluate hardware cost and speed.
Main Results:
- Simulations confirmed that the proposed neighborhood function enhances the vector quantization performance of hardware SOMs, even with the power-of-two restriction.
- FPGA implementation showed no increase in hardware cost or decrease in operating speed.
Conclusions:
- The novel neighborhood function effectively improves SOM vector quantization performance.
- It offers a hardware-efficient solution for accelerating SOMs, achieving high performance with parallel processing.
