Skip to main navigation Skip to search Skip to main content

Shallow Enough? A Cross-Architecture Study of Ultra-Low-Depth Neural Networks for Edge Inference

  • Chengwei Zhou
  • , Haotian Yu
  • , Shoma Yukawa
  • , Deniz Najafi
  • , Shaahin Angizi
  • , Gourav Datta

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Edge deployment imposes strict latency, memory, and energy constraints that scale directly with network depth, yet the question of which architectural paradigm offers the best accuracy-efficiency tradeoff at ultra-low depth (four to six layers) remains open. We present the first systematic comparison of convolutional, pure transformer, and hybrid CNN-transformer models under fixed shallow depth budgets. We develop an analytical framework characterizing the representational cost of local convolutional versus global attention-based feature mixing as a function of depth, and validate it with experiments on ImageNet-1K across architectures spanning MobileNetV2, DeiT, Swin, MobileViT, and EfficientFormer. Beyond FLOPs and accuracy, we report real-world latency and energy measurements on an NVIDIA Jetson Nano and examine deployment feasibility on Cortex-M class microcontrollers. Our experiments reveal how accuracy, latency, and memory scale across all three architecture families as depth decreases, providing practitioners with direct guidance on which model class to choose for a given depth and hardware budget.

Original languageEnglish (US)
Title of host publicationGLSVLSI 2026 - Proceedings of the Great Lakes Symposium on VLSI 2026
EditorsFan Chen, Peipei Zhou, Jie Gu, Amit R. Trivedi, Xiaoxuan Yang
PublisherAssociation for Computing Machinery, Inc
Pages220-225
Number of pages6
ISBN (Electronic)9798400724312
DOIs
StatePublished - Jun 22 2026
Event36th Great Lakes Symposium on VLSI, GLSVLSI 2026 - Canandaigua, United States
Duration: Jun 22 2026Jun 24 2026

Publication series

NameGLSVLSI 2026 - Proceedings of the Great Lakes Symposium on VLSI 2026

Conference

Conference36th Great Lakes Symposium on VLSI, GLSVLSI 2026
Country/TerritoryUnited States
CityCanandaigua
Period6/22/266/24/26

ASJC Scopus subject areas

  • Hardware and Architecture
  • Electrical and Electronic Engineering
  • Condensed Matter Physics

Keywords

  • CNN
  • Depth Reduction
  • Edge Inference
  • Hybrid Architectures
  • Vision Transformers

Cite this