Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Aksoy, Alperen, Bekman, Ilja, Dimitrov, Vesselin, Dorosti, Qader, Eguzo, Chimezie, Fleitmann, Sarah, Grewing, Christian, Hader, Fabian, Zambanini, Andre, van Waasen, Stefan
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2511.05479
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910244004691968
author Aksoy, Alperen
Bekman, Ilja
Dimitrov, Vesselin
Dorosti, Qader
Eguzo, Chimezie
Fleitmann, Sarah
Grewing, Christian
Hader, Fabian
Zambanini, Andre
van Waasen, Stefan
author_facet Aksoy, Alperen
Bekman, Ilja
Dimitrov, Vesselin
Dorosti, Qader
Eguzo, Chimezie
Fleitmann, Sarah
Grewing, Christian
Hader, Fabian
Zambanini, Andre
van Waasen, Stefan
contents This paper examines the use of Quantized Neural Networks (QNNs) for two resource-constrained scientific applications: automated calibration of semi-conductor quantum bits (qubits) and scientific particle detectors. We evaluate the trade-offs between Post-Training Quantization (PTQ), Quantization-Aware Training (QAT), and ultra-low-bit Binary Neural Networks (BNNs) with respect to latency and resource usage. Our results demonstrate that PTQ achieves a four-fold reduction in memory usage for U-shaped CNN (U-Net) architectures while maintaining or slightly enhancing segmentation accuracy (e.g. from 89% to 90% for a small U-Net with 447 parameters). For the training of non-differentiable custom BNNs , we propose a novel, hardware-constrained learning approach using Genetic Algorithms (GAs). We showcase a LUT-based BNN architecture suitable for direct conversion to VHDL via the HCL4BNN framework. This method achieves nanosecond-scale inference latencies (10-15 ns) without requiring specialized DSP or BRAM resources.
format Preprint
id arxiv_https___arxiv_org_abs_2511_05479
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Quantization Effects of Artificial Neural Networks for Embedded Edge-Computing Applications
Aksoy, Alperen
Bekman, Ilja
Dimitrov, Vesselin
Dorosti, Qader
Eguzo, Chimezie
Fleitmann, Sarah
Grewing, Christian
Hader, Fabian
Zambanini, Andre
van Waasen, Stefan
Neural and Evolutionary Computing
Instrumentation and Detectors
This paper examines the use of Quantized Neural Networks (QNNs) for two resource-constrained scientific applications: automated calibration of semi-conductor quantum bits (qubits) and scientific particle detectors. We evaluate the trade-offs between Post-Training Quantization (PTQ), Quantization-Aware Training (QAT), and ultra-low-bit Binary Neural Networks (BNNs) with respect to latency and resource usage. Our results demonstrate that PTQ achieves a four-fold reduction in memory usage for U-shaped CNN (U-Net) architectures while maintaining or slightly enhancing segmentation accuracy (e.g. from 89% to 90% for a small U-Net with 447 parameters). For the training of non-differentiable custom BNNs , we propose a novel, hardware-constrained learning approach using Genetic Algorithms (GAs). We showcase a LUT-based BNN architecture suitable for direct conversion to VHDL via the HCL4BNN framework. This method achieves nanosecond-scale inference latencies (10-15 ns) without requiring specialized DSP or BRAM resources.
title Quantization Effects of Artificial Neural Networks for Embedded Edge-Computing Applications
topic Neural and Evolutionary Computing
Instrumentation and Detectors
url https://arxiv.org/abs/2511.05479