随机记忆器的拓优化,用于输入意识的动态SNN
Bo Wang1,2, Xinyuan Zhang1,2, Shaocong Wang1,2
1Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
Science advances
|April 16, 2025
概括
我们开发了PRIME,这是一个以大脑为灵感的神经形态计算系统,使用记忆性尖端神经网络. PRIME显著提高了能源效率,并减少了AI任务的计算负载,模仿大脑功能.
科学领域:
- 神经形态工程的神经形态工程
- 人工智能的人工智能
- 材料科学 材料科学 材料科学
背景情况:
- 目前的机器学习模型,如GPT-4和SORA,缺乏人类大脑的效率和适应能力.
- 关键的局限性包括信号表示,优化,运行时重新配置和硬件架构的差异.
- ·诺伊曼瓶仍然是传统计算架构的一个重大挑战.
研究的目的:
- 推出一种新的脑启发的计算系统,即PRIME (用于输入意识的动态记忆尖端神经网络的修剪优化).
- 模拟大脑的尖端机制和结构可塑性,使用记忆尖端神经网络.
- 提高能源效率和减少人工智能硬件中的计算负载.
主要方法:
- 利用尖端神经元来模仿生物神经信号传递.
- 优化由结构可塑性启发的随机记忆尖端神经网络 (SNN) 的网络拓.
- 实施输入意识的早期停止政策,以减少处理延迟.
- 利用记忆式内存计算来克服·诺伊曼瓶.
主要成果:
- 在基于40nm,256-K内存晶体的宏观上,PRIME实现了与软件基线相匹配的分类准确度和开始得分.
- 显著提高了37.8×和62.5×的能源效率.
- 计算负载减少了77%和12.5%,性能降低最小.
- 展示了对随机记忆器噪声的强度.
结论:
- PRIME提供了一条通往高效,大脑启发的神经形态计算的可行途径.
- 该系统有效地减轻了与memristor编程随机性和·诺伊曼瓶相关的挑战.
- 在开发模拟人类大脑效率和适应能力的硬件方面,PRIME代表了重大进步.
相关概念视频
Biasing of P-N Junction
The operation of a p-n junction diode involves various biasing conditions, including forward bias, reverse bias, and equilibrium.
In equilibrium, no external voltage is applied across the p-n junction. The depletion region is formed at the junction interface due to the diffusion of carriers, which leaves behind charged dopants, acceptors on the p-side, and donors on the n-side. These immobile charges create an electric field that prevents further diffusion of carriers. The related energy band...
In equilibrium, no external voltage is applied across the p-n junction. The depletion region is formed at the junction interface due to the diffusion of carriers, which leaves behind charged dopants, acceptors on the p-side, and donors on the n-side. These immobile charges create an electric field that prevents further diffusion of carriers. The related energy band...
Biasing of Metal-Semiconductor Junctions
Biasing metal-semiconductor junctions involves applying a voltage across the junction. Specifically, the metal is connected to a voltage source, while the semiconductor is grounded. This technique is essential for controlling the direction and magnitude of current flow in electronic devices, including diodes, transistors, and photovoltaic cells.
In Schottky junctions, where the semiconductor is n-type, applying a positive voltage to the metal relative to the semiconductor reduces its Fermi...
In Schottky junctions, where the semiconductor is n-type, applying a positive voltage to the metal relative to the semiconductor reduces its Fermi...
MOS Capacitor
A Metal-Oxide-Semiconductor (MOS) capacitor is a fundamental structure used extensively in semiconductor device technology, particularly in the fabrication of integrated circuits and MOSFETs (metal-oxide-semiconductor field-effect transistors). The MOS capacitor consists of three layers: a metal gate, a dielectric oxide, and a semiconductor substrate.
The metal gate is typically made from highly conductive materials such as aluminum or polysilicon. Beneath the metal gate lies a thin layer of...
The metal gate is typically made from highly conductive materials such as aluminum or polysilicon. Beneath the metal gate lies a thin layer of...
MOSFET: Enhancement Mode
Enhancement-mode MOSFETs are pivotal components in electronics, distinguished by their capacity to act as highly efficient switches. They are part of the larger family of metal-oxide Semiconductor Field-Effect Transistors (MOSFETs). They are available in two types: p-channel and n-channel, each tailored to specific polarity operations.
In their basic form, enhancement-mode MOSFETs are typically non-conductive when the gate-source voltage (Vgs) is zero. This default 'off' state means no current...
In their basic form, enhancement-mode MOSFETs are typically non-conductive when the gate-source voltage (Vgs) is zero. This default 'off' state means no current...
Mechanistic Models: Compartment Models in Algorithms for Numerical Problem Solving
Mechanistic models play a crucial role in algorithms for numerical problem-solving, particularly in nonlinear mixed effects modeling (NMEM). These models aim to minimize specific objective functions by evaluating various parameter estimates, leading to the development of systematic algorithms. In some cases, linearization techniques approximate the model using linear equations.
In individual population analyses, different algorithms are employed, such as Cauchy's method, which uses a...
In individual population analyses, different algorithms are employed, such as Cauchy's method, which uses a...


