Abstract
An adaptive framework for edge devices is proposed herein that dynamically adjusts neural network precision (quantization) in real-time using live device telemetry. The framework ensures consistent accuracy, optimizes energy and thermal use, and seamlessly updates quantization, per layer, during inference with no downtime.
Publication Date
2026-01-04
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Calmîc, Sergiu, "ADAPTIVE ON-DEVICE QUANTIZATION CONTROLLER FOR DYNAMIC WORKLOADS", Technical Disclosure Commons, (January 04, 2026)
https://www.tdcommons.org/dpubs_series/9124
COinS