Efficient and Mathematically Robust Operations for Certified Neural Networks Inference
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916092735127552 |
|---|---|
| author | Geyer, Fabien Freitag, Johannes Schulz, Tobias Uhrig, Sascha |
| author_facet | Geyer, Fabien Freitag, Johannes Schulz, Tobias Uhrig, Sascha |
| contents | In recent years, machine learning (ML) and neural networks (NNs) have gained widespread use and attention across various domains, particularly in transportation for achieving autonomy, including the emergence of flying taxis for urban air mobility (UAM). However, concerns about certification have come up, compelling the development of standardized processes encompassing the entire ML and NN pipeline. This paper delves into the inference stage and the requisite hardware, highlighting the challenges associated with IEEE 754 floating-point arithmetic and proposing alternative number representations. By evaluating diverse summation and dot product algorithms, we aim to mitigate issues related to non-associativity. Additionally, our exploration of fixed-point arithmetic reveals its advantages over floating-point methods, demonstrating significant hardware efficiencies. Employing an empirical approach, we ascertain the optimal bit-width necessary to attain an acceptable level of accuracy, considering the inherent complexity of bit-width optimization. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_08225 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Efficient and Mathematically Robust Operations for Certified Neural Networks Inference Geyer, Fabien Freitag, Johannes Schulz, Tobias Uhrig, Sascha Machine Learning Hardware Architecture In recent years, machine learning (ML) and neural networks (NNs) have gained widespread use and attention across various domains, particularly in transportation for achieving autonomy, including the emergence of flying taxis for urban air mobility (UAM). However, concerns about certification have come up, compelling the development of standardized processes encompassing the entire ML and NN pipeline. This paper delves into the inference stage and the requisite hardware, highlighting the challenges associated with IEEE 754 floating-point arithmetic and proposing alternative number representations. By evaluating diverse summation and dot product algorithms, we aim to mitigate issues related to non-associativity. Additionally, our exploration of fixed-point arithmetic reveals its advantages over floating-point methods, demonstrating significant hardware efficiencies. Employing an empirical approach, we ascertain the optimal bit-width necessary to attain an acceptable level of accuracy, considering the inherent complexity of bit-width optimization. |
| title | Efficient and Mathematically Robust Operations for Certified Neural Networks Inference |
| topic | Machine Learning Hardware Architecture |
| url | https://arxiv.org/abs/2401.08225 |