### Abstract

The high computational and parameter complexity of neural networks makes their training very slow and difficult to deploy on energy and storage-constrained computing systems. Many network complexity reduction techniques have been proposed including fixed-point implementation. However, a systematic approach for designing full fixed-point training and inference of deep neural networks remains elusive. We describe a precision assignment methodology for neural network training in which all network parameters, i.e., activations and weights in the feedforward path, gradients and weight accumulators in the feedback path, are assigned close to minimal precision. The precision assignment is derived analytically and enables tracking the convergence behavior of the full precision training, known to converge a priori. Thus, our work leads to a systematic methodology of determining suitable precision for fixed-point training. The near optimality (minimality) of the resulting precision assignment is validated empirically for four networks on the CIFAR-10, CIFAR-100, and SVHN datasets. The complexity reduction arising from our approach is compared with other fixed-point neural network designs.

Original language | English (US) |
---|---|

State | Published - Jan 1 2019 |

Event | 7th International Conference on Learning Representations, ICLR 2019 - New Orleans, United States Duration: May 6 2019 → May 9 2019 |

### Conference

Conference | 7th International Conference on Learning Representations, ICLR 2019 |
---|---|

Country | United States |

City | New Orleans |

Period | 5/6/19 → 5/9/19 |

### Fingerprint

### ASJC Scopus subject areas

- Education
- Computer Science Applications
- Linguistics and Language
- Language and Linguistics

### Cite this

*Per-tensor fixed-point quantization of the back-propagation algorithm*. Paper presented at 7th International Conference on Learning Representations, ICLR 2019, New Orleans, United States.

**Per-tensor fixed-point quantization of the back-propagation algorithm.** / Sakr, Charbel; Shanbhag, Naresh R.

Research output: Contribution to conference › Paper

}

TY - CONF

T1 - Per-tensor fixed-point quantization of the back-propagation algorithm

AU - Sakr, Charbel

AU - Shanbhag, Naresh R

PY - 2019/1/1

Y1 - 2019/1/1

N2 - The high computational and parameter complexity of neural networks makes their training very slow and difficult to deploy on energy and storage-constrained computing systems. Many network complexity reduction techniques have been proposed including fixed-point implementation. However, a systematic approach for designing full fixed-point training and inference of deep neural networks remains elusive. We describe a precision assignment methodology for neural network training in which all network parameters, i.e., activations and weights in the feedforward path, gradients and weight accumulators in the feedback path, are assigned close to minimal precision. The precision assignment is derived analytically and enables tracking the convergence behavior of the full precision training, known to converge a priori. Thus, our work leads to a systematic methodology of determining suitable precision for fixed-point training. The near optimality (minimality) of the resulting precision assignment is validated empirically for four networks on the CIFAR-10, CIFAR-100, and SVHN datasets. The complexity reduction arising from our approach is compared with other fixed-point neural network designs.

AB - The high computational and parameter complexity of neural networks makes their training very slow and difficult to deploy on energy and storage-constrained computing systems. Many network complexity reduction techniques have been proposed including fixed-point implementation. However, a systematic approach for designing full fixed-point training and inference of deep neural networks remains elusive. We describe a precision assignment methodology for neural network training in which all network parameters, i.e., activations and weights in the feedforward path, gradients and weight accumulators in the feedback path, are assigned close to minimal precision. The precision assignment is derived analytically and enables tracking the convergence behavior of the full precision training, known to converge a priori. Thus, our work leads to a systematic methodology of determining suitable precision for fixed-point training. The near optimality (minimality) of the resulting precision assignment is validated empirically for four networks on the CIFAR-10, CIFAR-100, and SVHN datasets. The complexity reduction arising from our approach is compared with other fixed-point neural network designs.

UR - http://www.scopus.com/inward/record.url?scp=85071168236&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071168236&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85071168236

ER -