Towards an Expressivity-Normalized Energy-Demand Comparison of ANNs and SNNs
Mirrored from arXiv — Machine Learning for archival readability. Support the source by reading on the original site.
Computer Science > Machine Learning
Title:Towards an Expressivity-Normalized Energy-Demand Comparison of ANNs and SNNs
Abstract:Spiking neural networks (SNNs) are often regarded as energy-efficient alternatives to artificial neural networks (ANNs), yet their advantage depends critically on both network architecture and data properties. We develop an analytical framework to compare fully-connected ReLU ANNs and integrate-and-fire SNNs for time-series data with respect to their theoretical energy efficiency at matched expressive capacity. By relating an inference-energy model to theoretical bounds on representational expressivity, we derive an expressivity-normalized efficiency ratio and explicit thresholds in network width, spike sparsity, and ANN depth scaling. Our analysis characterizes the regimes in which event-driven computation offsets the temporal overhead of SNNs, providing capacity-aware principles for designing energy-efficient temporal networks. It shows that ANNs exceed SNNs in expressivity-normalized efficiency only in specific regimes.
| Subjects: | Machine Learning (cs.LG) |
| Cite as: | arXiv:2608.29869 [cs.LG] |
| (or arXiv:2608.29869v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2608.29869
arXiv-issued DOI via DataCite (pending registration)
|
Submission history
From: Miriam Kranzlmüller [view email][v1] Sun, 30 Aug 2026 15:56:17 UTC (100 KB)
Access Paper:
- View PDF
- HTML (experimental)
- TeX Source
References & Citations
Bookmark
Bibliographic and Citation Tools
Code, Data and Media Associated with this Article
Demos
Recommenders and Search Tools
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
More from arXiv — Machine Learning
-
M3-Former: Multimodal Transformer with Mixture-of-Experts for Long-Term Vessel Trajectory Prediction
Sep 11
-
Halo: Improving forecast accuracy through heteroscedastic estimation
Sep 11
-
Zero-shot rib design: merging training-free generative prior with topology optimization
Sep 11
-
Byzantine-Robust Federated Fire Detection with a Rotating Coordinator
Sep 11
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.