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

Neural Circuits01:25

Neural Circuits

Neural circuits and neuronal pools are two of the main structures found in the nervous system. Neural circuits are networks of neurons that work together to carry out a specific task or process. They consist of interconnected neurons and glial cells, which provide structural and metabolic support.
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...

You might also read

Related Articles

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

Sort by
Same author

Supernetwork-based efficient mapping of deep learning applications to mixed-precision hardware using model adaptation.

Nature communications·2026
Same author

Demonstration of transformer-based ALBERT model on a 14nm analog AI inference chip.

Nature communications·2025
Same author

Analog Resistive Switching Devices for Training Deep Neural Networks with the Novel Tiki-Taka Algorithm.

Nano letters·2024
Same author

Hardware-aware training for large-scale and diverse deep learning inference workloads using in-memory computing-based accelerators.

Nature communications·2023
Same author

Optimised weight programming for analogue memory-based deep neural networks.

Nature communications·2022
Same author

Autapses enhance bursting and coincidence detection in neocortical pyramidal cells.

Nature communications·2018
Same journal

Demonstration of a quantum C-NOT gate in a time-multiplexed fully reconfigurable photonic processor.

Nature communications·2026
Same journal

Nonlinear quantum light source with van der Waals ferroelectric NbOX<sub>2</sub> (X = Br, I).

Nature communications·2026
Same journal

Antagonistic histone H2A variants and autonomous heterochromatin formation shape epigenomic patterns in Arabidopsis.

Nature communications·2026
Same journal

The long tail of nitrate pollution in groundwater challenges governance of global water quality.

Nature communications·2026
Same journal

Select microbial metabolites promote tau aggregation in a murine tauopathy model.

Nature communications·2026
Same journal

Warming climate has lengthened global intense tropical cyclone seasons.

Nature communications·2026
See all related articles

Related Experiment Video

Updated: Jun 12, 2026

Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
11:18

Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks

Published on: March 2, 2015

10.3K

Fast and robust analog in-memory deep neural network training.

Malte J Rasch1,2, Fabio Carta3, Omobayode Fagbohungbe3

  • 1IBM Research, TJ Watson Research Center, Yorktown Heights, NY, USA. malte.rasch@gmail.com.

Nature Communications
|August 20, 2024
PubMed
Summary
This summary is machine-generated.

This study introduces improved analog in-memory training algorithms for deep learning, overcoming precision limitations and enabling faster, more robust neural network training without auxiliary digital computation.

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

9.0K
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

496

Related Experiment Videos

Last Updated: Jun 12, 2026

Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks
11:18

Closed-loop Neuro-robotic Experiments to Test Computational Properties of Neuronal Networks

Published on: March 2, 2015

10.3K
Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.0K
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

496

Area of Science:

  • Computer Science
  • Electrical Engineering
  • Materials Science

Background:

  • Analog in-memory computing offers efficient acceleration for deep learning networks.
  • Accelerating the inference phase is well-studied, but the more compute-intensive training phase is less explored.
  • Existing analog in-memory training algorithms face limitations such as significant digital compute overhead or strict programming requirements for zero points.

Purpose of the Study:

  • To propose novel algorithms for analog in-memory training that address the limitations of existing methods.
  • To achieve fast runtime complexity for in-memory training without requiring precise algorithmic zero points.
  • To investigate the impact of device non-idealities on algorithm performance.

Main Methods:

  • Development of two new algorithms for analog in-memory training.
  • Analysis of algorithm performance under various device non-idealities, including conductance noise, symmetry, retention, and endurance.
  • Comparison with existing approaches to highlight improvements in speed and precision requirements.

Main Results:

  • The proposed algorithms maintain fast runtime complexity for in-memory training.
  • The algorithms successfully resolve the need for precise programming of reference conductance values to establish an algorithmic zero point.
  • Investigation identified critical device material properties necessary for robust in-memory deep neural network training.

Conclusions:

  • The developed algorithms represent a significant advancement for analog in-memory deep learning training.
  • These algorithms reduce reliance on auxiliary digital computation and precise programming, enhancing efficiency.
  • Understanding device limitations is crucial for selecting appropriate materials for effective and reliable in-memory training hardware.