English(EN) Quadratic integrate-and-fire neurons exhibit less fragmented loss landscapes and outperform leaky integrate-and-fire neurons in spike-based gradient descent

二次神经元在基于脉冲的训练中优于泄漏神经元

作者 PulseAugur 编辑部 · [2 个来源] · 2026-06-02 17:26

研究人员已经证明，二次积分发放（QIF）神经元在训练脉冲神经网络方面比泄漏积分发放（LIF）神经元具有显著优势。通过在Spiking Heidelberg Digits数据集上进行的比较研究，QIF神经元表现出更优越的性能和更稳定的训练动态。该研究可视化了损失和梯度景观，揭示了LIF神经元由于不连续性而表现出碎片化和不稳定的梯度，而QIF神经元则提供了更平滑的训练体验。 AI

影响表明QIF神经元可以为神经形态计算应用实现更稳定有效的训练。

排序理由该集群包含一篇详细介绍神经网络研究新发现的学术论文。

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

arXiv cs.LG TIER_1 English(EN) · Carlo Wenig, Raoul-Martin Memmesheimer, Christian Klos · 2026-06-03 04:00

Quadratic integrate-and-fire neurons exhibit less fragmented loss landscapes and outperform leaky integrate-and-fire neurons in spike-based gradient descent

arXiv:2606.03935v1 Announce Type: cross Abstract: The ability to train spiking neural networks is essential for modeling biological neural networks as well as for neuromorphic computing. However, for the extensively used leaky integrate-and-fire (LIF) neurons, arbitrarily small p…
arXiv cs.NE (Neural & Evolutionary) TIER_1 English(EN) · Christian Klos · 2026-06-02 17:26

二次积分发放神经元展现出更少碎片化的损失景观，并在基于脉冲的梯度下降中优于泄漏积分发放神经元

The ability to train spiking neural networks is essential for modeling biological neural networks as well as for neuromorphic computing. However, for the extensively used leaky integrate-and-fire (LIF) neurons, arbitrarily small parameter changes can induce spike (dis)appearances…

报道来源 [2]

Quadratic integrate-and-fire neurons exhibit less fragmented loss landscapes and outperform leaky integrate-and-fire neurons in spike-based gradient descent

二次积分发放神经元展现出更少碎片化的损失景观，并在基于脉冲的梯度下降中优于泄漏积分发放神经元

相关实体

相关话题