Contrastive Self-Supervised Learning at the Edge: An Energy Perspective
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908585183674368 |
|---|---|
| author | Famá, Fernanda Pereira, Roberto Kalalas, Charalampos Dini, Paolo Qendro, Lorena Kawsar, Fahim Malekzadeh, Mohammad |
| author_facet | Famá, Fernanda Pereira, Roberto Kalalas, Charalampos Dini, Paolo Qendro, Lorena Kawsar, Fahim Malekzadeh, Mohammad |
| contents | While contrastive learning (CL) shows considerable promise in self-supervised representation learning, its deployment on resource-constrained devices remains largely underexplored. The substantial computational demands required for training conventional CL frameworks pose a set of challenges, particularly in terms of energy consumption, data availability, and memory usage. We conduct an evaluation of four widely used CL frameworks: SimCLR, MoCo, SimSiam, and Barlow Twins. We focus on the practical feasibility of these CL frameworks for edge and fog deployment, and introduce a systematic benchmarking strategy that includes energy profiling and reduced training data conditions. Our findings reveal that SimCLR, contrary to its perceived computational cost, demonstrates the lowest energy consumption across various data regimes. Finally, we also extend our analysis by evaluating lightweight neural architectures when paired with CL frameworks. Our study aims to provide insights into the resource implications of deploying CL in edge/fog environments with limited processing capabilities and opens several research directions for its future optimization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_08374 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Contrastive Self-Supervised Learning at the Edge: An Energy Perspective Famá, Fernanda Pereira, Roberto Kalalas, Charalampos Dini, Paolo Qendro, Lorena Kawsar, Fahim Malekzadeh, Mohammad Machine Learning While contrastive learning (CL) shows considerable promise in self-supervised representation learning, its deployment on resource-constrained devices remains largely underexplored. The substantial computational demands required for training conventional CL frameworks pose a set of challenges, particularly in terms of energy consumption, data availability, and memory usage. We conduct an evaluation of four widely used CL frameworks: SimCLR, MoCo, SimSiam, and Barlow Twins. We focus on the practical feasibility of these CL frameworks for edge and fog deployment, and introduce a systematic benchmarking strategy that includes energy profiling and reduced training data conditions. Our findings reveal that SimCLR, contrary to its perceived computational cost, demonstrates the lowest energy consumption across various data regimes. Finally, we also extend our analysis by evaluating lightweight neural architectures when paired with CL frameworks. Our study aims to provide insights into the resource implications of deploying CL in edge/fog environments with limited processing capabilities and opens several research directions for its future optimization. |
| title | Contrastive Self-Supervised Learning at the Edge: An Energy Perspective |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2510.08374 |