Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

Survival Tree01:19

Survival Tree

190
Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
 Building a Survival Tree
Constructing a...
190
Random and Systematic Errors01:20

Random and Systematic Errors

13.8K
Scientists always try their best to record measurements with the utmost accuracy and precision. However, sometimes errors do occur. These errors can be random or systematic. Random errors are observed due to the inconsistency or fluctuation in the measurement process, or variations in the quantity itself that is being measured. Such errors fluctuate from being greater than or less than the true value in repeated measurements. Consider a scientist measuring the length of an earthworm using a...
13.8K
Randomized Experiments01:13

Randomized Experiments

8.3K
The randomization process involves assigning study participants randomly to experimental or control groups based on their probability of being equally assigned. Randomization is meant to eliminate selection bias and balance known and unknown confounding factors so that the control group is similar to the treatment group as much as possible. A computer program and a random number generator can be used to assign participants to groups in a way that minimizes bias.
Simple randomization
Simple...
8.3K
Random Variables01:09

Random Variables

15.7K
A random variable is a single numerical value that indicates the outcome of a procedure. The concept of random variables is fundamental to the probability theory and was introduced by a Russian mathematician, Pafnuty Chebyshev, in the mid-nineteenth century.
Uppercase letters such as X or Y denote a random variable. Lowercase letters like x or y denote the value of a random variable. If X is a random variable, then X is written in words, and x is given as a number.
For example, let X = the...
15.7K

You might also read

Related Articles

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

Sort by
Same author

Secondary Electron Hyperspectral Imaging (SEHI) in FIB - SEM for Proton Exchange Membrane Fuel Cells Technology.

ACS omega·2026
Same author

Continuous Monitoring of Health and Mobility Indicators in Patients with Cardiovascular Disease: A Review of Recent Technologies.

Sensors (Basel, Switzerland)·2023
Same author

Traffic Estimation for Large Urban Road Network with High Missing Data Ratio.

Sensors (Basel, Switzerland)·2019
Same author

Robust Non-Rigid Feature Matching for Image Registration Using Geometry Preserving.

Sensors (Basel, Switzerland)·2019
Same author

A Review of Point Set Registration: From Pairwise Registration to Groupwise Registration.

Sensors (Basel, Switzerland)·2019

Related Experiment Video

Updated: Oct 27, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

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

Published on: December 15, 2023

722

Deep ConvNet: Non-Random Weight Initialization for Repeatable Determinism, Examined with FSGM.

Richard N M Rudd-Orthner1, Lyudmila Mihaylova1

  • 1Department of Automatic Control and Systems Engineering, University of Sheffield, Sheffield S1 3JD, UK.

Sensors (Basel, Switzerland)
|July 24, 2021
PubMed
Summary
This summary is machine-generated.

A novel non-random weight initialization method for neural networks accelerates learning and improves accuracy in image classification. This method demonstrates robustness and better retention of original data during transferred learning, even with significant distortions.

Keywords:
FSGMadversarial perturbation attackconvolutional layersmachine learningrepeatable determinismsmart sensorstransferred learningweight initialization

More Related Videos

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.5K

Related Experiment Videos

Last Updated: Oct 27, 2025

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
03:31

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

Published on: December 15, 2023

722
Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.5K

Area of Science:

  • Artificial Intelligence
  • Computer Vision
  • Machine Learning

Background:

  • Traditional weight initialization methods in neural networks rely on random values, which can impact learning efficiency.
  • Convolutional layers are crucial for image classification tasks, and their initialization significantly affects performance.
  • Transferred learning leverages pre-trained models but can be sensitive to dataset discrepancies and distortions.

Purpose of the Study:

  • To introduce and evaluate a deterministic, non-random weight initialization method for convolutional neural networks.
  • To assess the impact of this initialization on early learning and overall accuracy in image classification.
  • To investigate the method's robustness and data retention capabilities under controlled distortions in transferred learning scenarios.

Main Methods:

  • A repeatable, deterministic non-random weight initialization technique using number sequence substitution was developed.
  • The Fast Gradient Sign Method (FGSM) was employed to introduce controlled distortions and measure initialization effects.
  • The method was tested on benchmark models and a color image dataset (MTARSI2) with varying numerical similarities.

Main Results:

  • The proposed initialization method induced earlier learning, improving accuracy by 3-5% in the first epoch on a benchmark model.
  • Significant accuracy gains of ~10% were observed on the MTARSI2 dataset using a dissimilar model architecture.
  • The method showed robustness against optimization limitations like Glorot/Xavier and He initialization.
  • Under FGSM with high distortions (numerically dissimilar datasets), the method maintained ~31% accuracy compared to ~9% for standard approaches, indicating better original data retention.

Conclusions:

  • The deterministic non-random weight initialization method offers a promising alternative to random initialization for convolutional neural networks.
  • It enhances early learning and classification accuracy, particularly in transferred learning with challenging datasets.
  • The method demonstrates superior robustness and data retention properties compared to conventional initialization techniques.