Tunable Floating-Point for Embedded Machine Learning Algorithms Implementation

Marta Franceschi; Alberto Nannarelli; Maurizio Valle

doi:10.1109/SMACD.2018.8434873

Tunable Floating-Point for Embedded Machine Learning Algorithms Implementation

Marta Franceschi, Alberto Nannarelli, Maurizio Valle

University of Genoa

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Abstract

The development of embedded and real-time systems for Machine Learning data processing is challenging (e.g. IoT). Low latency, low power consumption and reduced hardware complexity should be the characteristics of such systems. Considering prosthetic applications, which are error-tolerant, a technique that tunes the precision of operands and operations has been chosen for a Machine Learning algorithm used for tactile data processing. This paper presents the implementation of a Tunable Floating-Point (TFP) representation into a Singular-Value Decomposition (SVD) algorithm based on the One-Sided Jacobi method. The TFP representations demonstrate high performance and efficiency improvements of the SVD algorithm.

Original language	English
Title of host publication	Proceedings of 2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design
Publisher	IEEE
Publication date	2018
Pages	89-92
ISBN (Print)	9781538651520
DOIs	https://doi.org/10.1109/SMACD.2018.8434873
Publication status	Published - 2018
Event	2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design - Czech Technical University in Prague, Prague, Czech Republic Duration: 2 Jul 2018 → 5 Jul 2018 Conference number: 15

Conference

Conference	2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design
Number	15
Location	Czech Technical University in Prague
Country/Territory	Czech Republic
City	Prague
Period	02/07/2018 → 05/07/2018

Keywords

Simulation
Power efficiency
SVD algorithm

Access to Document

10.1109/SMACD.2018.8434873

OpenUrl availability

Full text

Cite this

@inproceedings{5a7087ef78514a2a9dbadd194d41d2bf,

title = "Tunable Floating-Point for Embedded Machine Learning Algorithms Implementation",

abstract = "The development of embedded and real-time systems for Machine Learning data processing is challenging (e.g. IoT). Low latency, low power consumption and reduced hardware complexity should be the characteristics of such systems. Considering prosthetic applications, which are error-tolerant, a technique that tunes the precision of operands and operations has been chosen for a Machine Learning algorithm used for tactile data processing. This paper presents the implementation of a Tunable Floating-Point (TFP) representation into a Singular-Value Decomposition (SVD) algorithm based on the One-Sided Jacobi method. The TFP representations demonstrate high performance and efficiency improvements of the SVD algorithm.",

keywords = "Simulation, Power efficiency, SVD algorithm",

author = "Marta Franceschi and Alberto Nannarelli and Maurizio Valle",

year = "2018",

doi = "10.1109/SMACD.2018.8434873",

language = "English",

isbn = "9781538651520 ",

pages = "89--92",

booktitle = "Proceedings of 2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design",

publisher = "IEEE",

address = "United States",

note = "2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, SMACD 2018 ; Conference date: 02-07-2018 Through 05-07-2018",

}

Franceschi, M, Nannarelli, A & Valle, M 2018, Tunable Floating-Point for Embedded Machine Learning Algorithms Implementation. in Proceedings of 2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design. IEEE, pp. 89-92, 2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, Prague, Czech Republic, 02/07/2018. https://doi.org/10.1109/SMACD.2018.8434873

Tunable Floating-Point for Embedded Machine Learning Algorithms Implementation. / Franceschi, Marta; Nannarelli, Alberto; Valle, Maurizio.
Proceedings of 2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design. IEEE, 2018. p. 89-92.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

TY - GEN

T1 - Tunable Floating-Point for Embedded Machine Learning Algorithms Implementation

AU - Franceschi, Marta

AU - Nannarelli, Alberto

AU - Valle, Maurizio

N1 - Conference code: 15

PY - 2018

Y1 - 2018

N2 - The development of embedded and real-time systems for Machine Learning data processing is challenging (e.g. IoT). Low latency, low power consumption and reduced hardware complexity should be the characteristics of such systems. Considering prosthetic applications, which are error-tolerant, a technique that tunes the precision of operands and operations has been chosen for a Machine Learning algorithm used for tactile data processing. This paper presents the implementation of a Tunable Floating-Point (TFP) representation into a Singular-Value Decomposition (SVD) algorithm based on the One-Sided Jacobi method. The TFP representations demonstrate high performance and efficiency improvements of the SVD algorithm.

AB - The development of embedded and real-time systems for Machine Learning data processing is challenging (e.g. IoT). Low latency, low power consumption and reduced hardware complexity should be the characteristics of such systems. Considering prosthetic applications, which are error-tolerant, a technique that tunes the precision of operands and operations has been chosen for a Machine Learning algorithm used for tactile data processing. This paper presents the implementation of a Tunable Floating-Point (TFP) representation into a Singular-Value Decomposition (SVD) algorithm based on the One-Sided Jacobi method. The TFP representations demonstrate high performance and efficiency improvements of the SVD algorithm.

KW - Simulation

KW - Power efficiency

KW - SVD algorithm

U2 - 10.1109/SMACD.2018.8434873

DO - 10.1109/SMACD.2018.8434873

M3 - Article in proceedings

SN - 9781538651520

SP - 89

EP - 92

BT - Proceedings of 2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design

PB - IEEE

T2 - 2018 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design

Y2 - 2 July 2018 through 5 July 2018

ER -

Tunable Floating-Point for Embedded Machine Learning Algorithms Implementation

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this