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: Jun 11, 2025

Deep Neural Networks for Image-Based Dietary Assessment

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

MixTrain: accelerating DNN training via input mixing.

Sarada Krithivasan¹, Sanchari Sen², Swagath Venkataramani²

¹Department of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, United States.

Frontiers in Artificial Intelligence

|October 2, 2024

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

Who needs an electrical back-contact after all?

Nanoscale·2026

Same author

3D Printing and Guided Endodontics: The Future of Precision Access Cavity Design.

Cureus·2025

Same author

LRMP: Layer Replication with Mixed Precision for spatial in-memory DNN accelerators.

Frontiers in artificial intelligence·2024

Same author

RoMIA: a framework for creating Robust Medical Imaging AI models for chest radiographs.

Frontiers in radiology·2024

Same author

Compute in-Memory with Non-Volatile Elements for Neural Networks: A Review from a Co-Design Perspective.

Advanced materials (Deerfield Beach, Fla.)·2022

Same author

Accelerating DNN Training Through Selective Localized Learning.

Frontiers in neuroscience·2022

Same journal

Cross-linguistic patterns of cognitive biases in large language models: a comparative study in English, Hebrew, and Russian.

Frontiers in artificial intelligence·2026

Same journal

From human-like AI to user adoption: the role of trust, attitude, and social influence in shaping behavioral intention.

Frontiers in artificial intelligence·2026

Same journal

Building large-scale English-Romanian literary translation resources with open models.

Frontiers in artificial intelligence·2026

Same journal

Logic, inference, understanding: cross-domain generalization for generative language models.

Frontiers in artificial intelligence·2026

Same journal

Label tree semantic losses for rich multi-class medical image segmentation.

Frontiers in artificial intelligence·2026

Same journal

Score-based generative diffusion models to synthesize full-dose FDG brain PET from MRI in epilepsy patients.

Frontiers in artificial intelligence·2026

See all related articles

Input mixing combines multiple data inputs into one for training Deep Neural Networks (DNNs), significantly reducing training time. This method speeds up DNN training by processing fewer mini-batches per epoch.

Area of Science:

Computer Science
Artificial Intelligence
Machine Learning

Background:

Deep Neural Networks (DNNs) training demands substantial computational resources, time, and energy.
Increasing dataset complexity is a key factor in the lengthy training times for state-of-the-art DNNs.

Purpose of the Study:

To investigate input mixing as a strategy to accelerate DNN training.
To mitigate accuracy degradation caused by naive input mixing.

Main Methods:

Developed input mixing by combining multiple inputs into a single composite input with a composite label.
Introduced strategies to reduce inter-input interference by exploiting feature spatial separation and adaptively varying mixing ratios.
Proposed heuristics for automatic selection of data subsets for mixing.

Keywords:

GPUs (graphics processing units)deep learning input mixing runtime efficiency training

More Related Videos

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

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: Jun 11, 2025

Deep Neural Networks for Image-Based Dietary Assessment

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

Main Results:

Achieved up to 1.6x and 1.8x training speedups on ImageNet and Cifar10 datasets, respectively.
Demonstrated speedups across various architectures including ResNets, MobileNetV2, and Vision Transformers.
Maintained negligible loss in classification accuracy.

Conclusions:

Input mixing, when optimized with proposed strategies, can significantly reduce DNN training time.
The developed methods enable faster training of deep learning models without compromising performance.
This approach offers a practical solution for computationally intensive deep learning tasks.