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

Feedback Inhibition00:46

Feedback Inhibition

56.9K
Biochemical reactions are occurring constantly in cells, converting starting substances to different products, usually with the help of enzymes that speed the reactions. Without enzymes, it would take far too long for most reactions to occur to be useful to the cell!
56.9K
Feedback Loops01:01

Feedback Loops

64.2K
In most cases, excessive hormone production is prevented by negative feedback—a loop that starts with a stimulus inducing the release of a particular substance, like a hormone, to maintain a certain level before triggering a signal that results in a decrease in further release of the hormone.
64.2K
Effects of feedback01:24

Effects of feedback

1.0K
Feedback in control systems plays a critical role in shaping various operational parameters, extending beyond simple error reduction to influence stability, bandwidth, gain, impedance, and sensitivity. Understanding these effects requires examining a basic feedback system characterized by defined input, output, error, and feedback signals.
Feedback significantly modifies the gain of a control system. The gain of a system without feedback is altered by a factor of one plus GH, where G represents...
1.0K
Feedback control systems01:26

Feedback control systems

696
Feedback control systems are categorized in various ways based on their design, analysis, and signal types.
Linear feedback systems are theoretical models that simplify analysis and design. These systems operate under the principle that their output is directly proportional to their input within certain ranges. For instance, an amplifier in a control system behaves linearly as long as the input signal remains within a specific range. However, most physical systems exhibit inherent nonlinearity...
696
Facial Feedback Hypothesis01:24

Facial Feedback Hypothesis

619
Charles Darwin proposed that facial expressions are an evolutionary adaptation for communication. He argued that these expressions are not influenced by culture but are universal across species. For example, a snarling expression with exposed teeth signals a threat in many animals, including humans. Darwin also suggested that displaying an emotion can intensify the feeling. Smiling, for example, could enhance one's sense of happiness. This idea laid the foundation for understanding the role...
619
Cell Signaling Feedback Loops01:07

Cell Signaling Feedback Loops

7.3K
Positive and negative feedback loops are crucial for regulating biological signaling systems. These feedback loops are processes that connect output signals to their inputs.
Negative feedback loops
Most signaling systems have negative feedback loops that can perform different functions such as output limiter, and adaptation.
Output limiter
Upon receiving an input signal, the cellular response rapidly increases until a threshold is reached. Beyond this threshold, a negative feedback loop...
7.3K

You might also read

Related Articles

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

Sort by
Same author

A wireless subdural-contained brain-computer interface with 65,536 electrodes and 1,024 channels.

Nature electronics·2026
Same author

Risk Factors for Urinary Tract Infections in Patients with Multiple Sclerosis.

Neurology India·2026
Same author

Stable, chronic in-vivo recordings from a fully wireless subdural-contained 65,536-electrode brain-computer interface device.

bioRxiv : the preprint server for biology·2024
Same author

E-Gaze: Gaze Estimation With Event Camera.

IEEE transactions on pattern analysis and machine intelligence·2024
Same author

Bio-mimetic high-speed target localization with fused frame and event vision for edge application.

Frontiers in neuroscience·2022
Same author

Roadmap on emerging hardware and technology for machine learning.

Nanotechnology·2020

Related Experiment Video

Updated: Jan 23, 2026

Control of Eating Behavior Using a Novel Feedback System
04:48

Control of Eating Behavior Using a Novel Feedback System

Published on: May 8, 2018

11.5K

Direct Feedback Alignment With Sparse Connections for Local Learning.

Brian Crafton1, Abhinav Parihar1, Evan Gebhardt1

  • 1School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA, United States.

Frontiers in Neuroscience
|June 11, 2019
PubMed
Summary
This summary is machine-generated.

This study introduces a bio-plausible alternative to backpropagation for training deep neural networks (DNNs). The new method significantly reduces data movement and improves computational efficiency, offering a more scalable solution for machine learning hardware.

Keywords:
backpropagationbio-plausible algorithmsfeedback alignmenthardware accelerationlocal learningsparse neural networks

More Related Videos

Movement Retraining using Real-time Feedback of Performance
08:16

Movement Retraining using Real-time Feedback of Performance

Published on: January 17, 2013

13.8K
A Vibrotactile Feedback Device for Seated Balance Assessment and Training
09:13

A Vibrotactile Feedback Device for Seated Balance Assessment and Training

Published on: January 20, 2019

6.8K

Related Experiment Videos

Last Updated: Jan 23, 2026

Control of Eating Behavior Using a Novel Feedback System
04:48

Control of Eating Behavior Using a Novel Feedback System

Published on: May 8, 2018

11.5K
Movement Retraining using Real-time Feedback of Performance
08:16

Movement Retraining using Real-time Feedback of Performance

Published on: January 17, 2013

13.8K
A Vibrotactile Feedback Device for Seated Balance Assessment and Training
09:13

A Vibrotactile Feedback Device for Seated Balance Assessment and Training

Published on: January 20, 2019

6.8K

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computational Neuroscience

Background:

  • Deep neural networks (DNNs) rely on backpropagation and gradient descent for training, which face efficiency challenges in large-scale networks due to the weight transport problem.
  • The weight transport problem necessitates extensive data movement, hindering performance and energy efficiency in machine learning hardware.

Purpose of the Study:

  • To propose a bio-plausible alternative to backpropagation that addresses the weight transport problem.
  • To enhance the performance and energy efficiency of machine learning hardware by reducing data movement.

Main Methods:

  • Developed a novel algorithm based on feedback alignment principles, reducing error computation at each synapse to a product of three scalar values.
  • Utilized a sparse feedback matrix, enabling neurons to require only a fraction of the information used in prior feedback alignment methods.
  • Evaluated the algorithm on standard datasets, including ImageNet, to assess scalability for complex problems.

Main Results:

  • Achieved orders of magnitude improvement in data movement and a 2x improvement in multiply-and-accumulate operations compared to backpropagation.
  • Demonstrated that direct feedback alignment, when applied to fully connected layers after transferring trained convolutional layers, yields results competitive with backpropagation.
  • Observed that using an extremely sparse feedback matrix results in minimal accuracy loss while providing significant hardware advantages.

Conclusions:

  • The proposed feedback alignment algorithm offers a more efficient and scalable approach to training deep neural networks, particularly for hardware implementations.
  • Direct feedback alignment, combined with strategic layer training, can achieve performance comparable to backpropagation with enhanced hardware efficiency.
  • Sparse feedback matrices are a viable strategy for further optimizing hardware advantages in feedback alignment methods without substantial accuracy degradation.