对于强大的尖端神经网络训练而言,生物知情的刺激和抑制比率
Joseph A Kilgore1, Jeffrey D Kopsick2, Giorgio A Ascoli2
1Department of Electrical and Computer Engineering, George Washington University, Washington, 20052, USA.
Scientific reports
|July 9, 2025
概括
为节能人工智能培养尖端神经网络 (SNN) 是一个挑战. 这项研究确定了关键因素,如低发射率和抑制模式,使SNN训练能够强大,特别是具有生物现实的神经元比率.
科学领域:
- 计算神经科学是一种计算神经科学.
- 人工智能的人工智能是人工智能.
- 神经形态工程的神经形态工程
背景情况:
- 尖端神经网络 (SNN) 通过模仿生物大脑,提供节能的人工智能.
- 训练SNN,特别是对激发性和抑制性连接的生物学约束,提出了重大挑战.
- 强有力的SNN培训原则对于其实际应用至关重要.
研究的目的:
- 确定影响不同激发性-抑制性 (E:I) 神经元比率的SNNs可训练性的关键因素.
- 调查生物约束的影响,如发射率和抑制性尖端模式,SNN培训.
- 评估训练有素的SNN在杂环境中的强度及其对抑制神经元功能的依赖.
主要方法:
- 模拟的尖端神经网络,具有不同的EI比率.
- 分析不同初始发射速率和抑制性增强模式下的训练动态.
- 利用范罗斯姆距离来量化尖峰列车同步和网络稳定性.
主要成果:
- 低初始发射速率和多种抑制性尖端模式对于成功的SNN训练至关重要.
- 生物现实的E:I比率使得即使在低活动水平和噪音条件下,也能提供可靠的培训.
- 抑制性神经元显著提高网络对噪声的稳定性,正如范罗斯姆距离分析所示.
结论:
- 确定了生物约束性SNNs的关键培训原则.
- 证明了现实的E:I比率和特定的抑制动力学可以提高训练的可靠性和稳定性.
- 这些发现支持开发生物信息的大规模SNNs和节能硬件.
更多相关视频
07:51Inhibitory Synapse Formation in a Co-culture Model Incorporating GABAergic Medium Spiny Neurons and HEK293 Cells Stably Expressing GABAA Receptors
Published on: November 14, 2014
17.5K
08:08Real-time Electrophysiology: Using Closed-loop Protocols to Probe Neuronal Dynamics and Beyond
Published on: June 24, 2015
11.6K
相关概念视频
01:29Excitatory and Inhibitory Effects of Neurotransmitters
10.8K
When an action potential reaches the presynaptic axon terminal, it releases neurotransmitters from the neuron into the synaptic cleft at a chemical synapse. The released neurotransmitter can be excitatory or inhibitory. The critical criteria commonly used to determine whether a molecule is a neurotransmitter at a chemical synapse are the molecule's presence in the presynaptic neuron. Second, its release is in response to strong presynaptic depolarization. And lastly, the presence of...
10.8K
01:19The Role of Ion Channels in Neuronal Computation
3.3K
A postsynaptic neuron usually receives numerous impulses from several other presynaptic neurons. The axon hillock of the postsynaptic neuron integrates all these signals and determines the likelihood of firing an action potential.
Sometimes a single EPSP is strong enough to induce an action potential in the postsynaptic neuron. However, multiple presynaptic inputs must often create EPSPs around the same time for the postsynaptic neuron to be sufficiently depolarized to fire an action potential....
Sometimes a single EPSP is strong enough to induce an action potential in the postsynaptic neuron. However, multiple presynaptic inputs must often create EPSPs around the same time for the postsynaptic neuron to be sufficiently depolarized to fire an action potential....
3.3K