Related Experiment Video
Updated: May 14, 2026
10:50
Computational Modeling of Retinal Neurons for Visual Prosthesis Research - Fundamental Approaches
Published on: June 21, 2022
Resistive memory-based neural differential equation solver for score-based diffusion model
Jichang Yang1,2,3, Hegan Chen1,2,3, Jia Chen4
1Department of Electrical and Computer Engineering, The University of Hong Kong, Hong Kong, China.
Nature Communications
|May 12, 2026
Summary
This study introduces a novel brain-inspired analog in-memory computing system for faster and more energy-efficient AI content generation. The new hardware significantly accelerates generative AI tasks while reducing power consumption compared to digital systems.
Area of Science:
- Artificial Intelligence
- Computer Engineering
- Neuroscience
Background:
- Current AI-generated content (AIGC) models, like score-based diffusion, are computationally expensive and slow.
- Digital computers face bottlenecks due to separate storage and processing, hindering continuous generation dynamics.
Purpose of the Study:
- To develop a more efficient hardware solution for generative AI.
- To overcome the limitations of digital computing in AI content generation.
Main Methods:
- Implemented a brain-inspired analog in-memory computing system using resistive memory.
- Integrated storage and computation for time-continuous neural differential equation solving.
- Experimentally validated the system using 180 nm resistive memory macros.
Main Results:
- Achieved significant acceleration in unconditional (69.0x) and conditional (116.5x) generation tasks compared to digital hardware.
- Reduced energy consumption by 31.5% and 52.0% for unconditional and conditional generation, respectively.
- Maintained generative quality equivalent to software baselines.
Conclusions:
- The analog in-memory computing system offers a promising hardware acceleration for generative AI.
- This approach enhances efficiency for edge computing applications in generative AI.
- Brain-inspired computing architectures can overcome limitations of conventional digital systems for AI tasks.
Related Concept Videos
01:25Modeling with Differential Equations
Population dynamics can be described mathematically by considering the population size P(t) as a function of time. The rate of change of the population is then represented by the derivative of P(t). A simple assumption is that the rate of growth is proportional to the size of the population itself. This leads to an exponential growth model, where the population increases rapidly without bound. While this is a useful first approximation, it does not reflect realistic long-term...
01:21Differential Equations: Problem Solving
When analyzing the motion of falling objects, it is essential to consider not only the force of gravity but also the opposing force of air resistance. A practical example involves releasing a heavy test weight during a safety check on a ship. As the weight falls from rest, gravity accelerates it downward while air resistance exerts an upward force that increases with velocity. This dynamic interplay of forces is well described by differential equations, which provide a mathematical framework...
01:27Linear Differential Equations
The integrating factor method provides a systematic way to solve first-order linear differential equations, especially those that cannot be handled by separation of variables. This method is particularly useful in modeling time-dependent physical systems influenced by both constant inputs and resistive forces. A common example is the motion of a car subjected to a constant engine force while experiencing air resistance proportional to its velocity.In such scenarios, Newton’s second law yields a...
01:20Separable Differential Equations
A separable differential equation is a type of first-order differential equation where the derivative dy/dx can be expressed as a product of two functions: one that depends only on x and another that depends only on y. This allows for the rearrangement of the equation so that all terms involving y are on one side, and all terms involving x are on the other. This process, known as the separation of variables, simplifies the process of solving the equation by enabling the integration of both...
01:20Introduction to Differential Equations
A differential equation is a mathematical expression that establishes a relationship between a function and its derivatives. These equations are fundamental in modeling dynamic systems across various fields of science and engineering. The order of a differential equation is defined by the highest order derivative present in the equation. A first-order differential equation includes only the first derivative, while a second-order differential equation includes up to the second derivative of the...
00:57Physiological Pharmacokinetic Models: Blood Flow-Limited Versus Diffusion-Limited Models
Physiological pharmacokinetic models, often called flow-limited or perfusion models, typically assume a swift drug distribution between tissue and venous blood, creating a rapid drug equilibrium. This premise is based on the idea that drug diffusion is extremely fast, and the cell membrane presents no barrier to drug permeation. In this scenario, where no drug binding occurs, the drug concentration in the tissue equals that of the venous blood leaving the tissue. This greatly simplifies the...