Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Jan 13, 2026

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Image generator for tabular data based on non-Euclidean metrics for CNN-based classification.

Yu-Rong Lin¹, Han-Ming Wu¹

¹Department of Statistics, National Chengchi University, Taipei City, Taiwan, Republic of China.

|January 9, 2026

Summary

This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Quantum gyroscope based on the cavity-magnon system.

Optics express·2026

Same author

PSA density improves the rate of prostate cancer detection in Chinese men with a PSA between 2.5-10.0 ng ml (-1) and 10.1-20.0 ng ml (-1) : a multicenter study.

Asian journal of andrology·2014

Same author

[2-deoxyglucose enhances chemosensitivity of androgen-independent prostate cancer cells to docetaxel].

Zhonghua yi xue za zhi·2014

Same author

Nanoscale dynamic behavior of surface magnetic domains on a La0.7Sr0.3MnO3 thin film.

Journal of nanoscience and nanotechnology·2013

Same author

Prostate volume as an independent predictor of prostate cancer in men with PSA of 10-50 ng ml(-1).

Asian journal of andrology·2013

Same author

A nomogram based on age, prostate-specific antigen level, prostate volume and digital rectal examination for predicting risk of prostate cancer.

Asian journal of andrology·2013

Same journal

Epidemiological characteristics of amebiasis in Japan from 2001 to 2022.

PloS one·2026

Same journal

Longitudinal associations of academic stress with eating related patterns, nutrition, somatic indicators, and depressive symptoms in university students: A study protocol.

PloS one·2026

Same journal

Pollution removal efficiency enhancement by agricultural biomass additions in constructed wetlands: A framework integrating meta-analysis with explainable machine learning.

PloS one·2026

Same journal

Insulation failure mapping on power transformer bushing using FRA and electrostatic simulation.

PloS one·2026

Same journal

Enhancing medical Q&A systems with multimodal knowledge graphs and dual-layer attention mechanisms.

PloS one·2026

Same journal

UAMP: Consistent video object segmentation with uncertainty-aware memory propagation.

PloS one·2026

See all related articles

This study enhances deep learning for tabular data by using non-Euclidean distances in the Image Generator for Tabular Data (IGTD) framework. This improves Convolutional Neural Network (CNN) classification accuracy on complex datasets.

Area of Science:

Machine Learning
Data Science
Computational Biology

Background:

Tabular data is crucial for analysis but challenging for conventional methods due to high dimensionality and nonlinear relationships.
Deep learning models like Convolutional Neural Networks (CNNs) excel at feature extraction but are typically used for image data.
The Image Generator for Tabular Data (IGTD) framework converts tabular data into images for CNN analysis, originally using Euclidean distance.

Purpose of the Study:

To evaluate the effectiveness of non-Euclidean distance metrics within the IGTD framework for CNN-based classification of tabular data.
To compare alternative distance metrics (e.g., Jensen-Shannon, Wasserstein) against Euclidean distance for capturing complex feature relationships.
To assess the impact of these metrics on classification accuracy and the structural fidelity of generated image representations.

More Related Videos

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

Related Experiment Videos

Last Updated: Jan 13, 2026

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma

Published on: October 10, 2018

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

Main Methods:

Implemented the IGTD framework with several non-Euclidean distance metrics: one minus correlation, Geodesic, Jensen-Shannon, Wasserstein, and Tropical distance.
Conducted comparative experiments using both simulated and real-world genomics datasets.
Evaluated performance based on classification accuracy and the structural integrity of the image representations generated from tabular data.

Main Results:

Non-Euclidean distance metrics significantly enhanced the performance of CNN-based classification on tabular data compared to the standard Euclidean distance.
The chosen non-Euclidean metrics demonstrated a superior ability to capture intricate and nonlinear relationships among data features.
Improved classification accuracy was observed across various datasets, indicating the broader applicability of the enhanced IGTD framework.

Conclusions:

The integration of non-Euclidean distance metrics into the IGTD framework offers a powerful approach for applying CNNs to high-dimensional tabular data.
This method provides a flexible and interpretable solution for complex data analysis across diverse scientific domains.
The findings suggest that adapting distance metrics is key to unlocking the full potential of deep learning for structured, non-image data.