Running Conventional Automatic Speech Recognition on Memristor Hardware: A Simulated Approach

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Rossenbach, Nick, Hilmes, Benedikt, Brackmann, Leon, Gunz, Moritz, Schlüter, Ralf
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913867851890688
author Rossenbach, Nick
Hilmes, Benedikt
Brackmann, Leon
Gunz, Moritz
Schlüter, Ralf
author_facet Rossenbach, Nick
Hilmes, Benedikt
Brackmann, Leon
Gunz, Moritz
Schlüter, Ralf
contents Memristor-based hardware offers new possibilities for energy-efficient machine learning (ML) by providing analog in-memory matrix multiplication. Current hardware prototypes cannot fit large neural networks, and related literature covers only small ML models for tasks like MNIST or single word recognition. Simulation can be used to explore how hardware properties affect larger models, but existing software assumes simplified hardware. We propose a PyTorch-based library based on "Synaptogen" to simulate neural network execution with accurately captured memristor hardware properties. For the first time, we show how an ML system with millions of parameters would behave on memristor hardware, using a Conformer trained on the speech recognition task TED-LIUMv2 as example. With adjusted quantization-aware training, we limit the relative degradation in word error rate to 25% when using a 3-bit weight precision to execute linear operations via simulated analog computation.
format Preprint
id arxiv_https___arxiv_org_abs_2505_24721
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Running Conventional Automatic Speech Recognition on Memristor Hardware: A Simulated Approach
Rossenbach, Nick
Hilmes, Benedikt
Brackmann, Leon
Gunz, Moritz
Schlüter, Ralf
Machine Learning
Hardware Architecture
Emerging Technologies
Memristor-based hardware offers new possibilities for energy-efficient machine learning (ML) by providing analog in-memory matrix multiplication. Current hardware prototypes cannot fit large neural networks, and related literature covers only small ML models for tasks like MNIST or single word recognition. Simulation can be used to explore how hardware properties affect larger models, but existing software assumes simplified hardware. We propose a PyTorch-based library based on "Synaptogen" to simulate neural network execution with accurately captured memristor hardware properties. For the first time, we show how an ML system with millions of parameters would behave on memristor hardware, using a Conformer trained on the speech recognition task TED-LIUMv2 as example. With adjusted quantization-aware training, we limit the relative degradation in word error rate to 25% when using a 3-bit weight precision to execute linear operations via simulated analog computation.
title Running Conventional Automatic Speech Recognition on Memristor Hardware: A Simulated Approach
topic Machine Learning
Hardware Architecture
Emerging Technologies
url https://arxiv.org/abs/2505.24721