arXiv：cs.LG· Kohei Tsuchiyama, Andre Roehm, Takatomo Mihana, Ryoichi Horisaki·· 7 小时前AI 评分28
TIDAL-Net：面向物理神经网络的时间复用层重用架构
Time-multiplexed layer reuse for physical neural networks
AI 导读
针对物理神经网络（PNN）权重调整缓慢的瓶颈，研究者提出 TIDAL-Net，一种介于循环与深度网络之间的时间索引深度交替层网络。它利用 PNN 中快速前向动力学与缓慢可训练权重之间的时间尺度分离，通过逐层时间复用增加有效深度并控制实现成本。在图像分类和自然语言处理任务上的数值实验显示，仅需对传统 PNN 做少量修改，TIDAL-Net 即可提升性能。
正文
Abstract:Physical neural networks (PNNs) are promising candidates for next-generation computing, but existing demonstrations remain several orders of magnitude smaller than modern digital neural networks, whose recent advances have been driven by rapid growth in trainable parameters. This situation resembles the constraints of early digital neural networks, which led to ideas around parameter reuse. We investigate what similarly efficient hardware architectures may look like, focusing specifically on the common bottleneck of slow re-adjustment of the weights in PNNs. We propose the Time-Indexed Deep Alternating Layers Network (TIDAL-Net), which occupies an intermediate regime between recurrent and deep neural networks, specifically aimed at the scales and restrictions of common PNN prototypes. TIDAL-Net leverages the timescale separation found in many PNNs between fast forward dynamics and slowly trainable weights and biases, using layer-by-layer time multiplexing to increase effective depth while limiting implementation cost. Numerical experiments on image classification and natural language processing tasks show that TIDAL-Net improves performance with only minor modifications to conventional PNNs.
| Subjects: | Machine Learning (cs.LG); Adaptation and Self-Organizing Systems (nlin.AO) |
| Cite as: | arXiv:2511.00044 [cs.LG] |
| (or arXiv:2511.00044v4 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2511.00044 arXiv-issued DOI via DataCite |
Submission history
From: Kohei Tsuchiyama [view email]
[v1]
Tue, 28 Oct 2025 07:25:41 UTC (2,530 KB)
[v2]
Tue, 18 Nov 2025 04:13:18 UTC (1,457 KB)
[v3]
Wed, 10 Jun 2026 09:24:32 UTC (1,401 KB)
[v4]
Tue, 6 Oct 2026 09:33:14 UTC (1,665 KB)
来源：arXiv：cs.LG · arxiv.org