FAPNN: An FPGA based Approximate Probabilistic Neural Network Library

Kizheppatt Vipin; Yerbol Akhmetov; Serikbolsyn Myrzakhme; Alex P. James

doi:10.1109/CoCoNet.2018.8476889

FAPNN: An FPGA based Approximate Probabilistic Neural Network Library

Kizheppatt Vipin, Yerbol Akhmetov, Serikbolsyn Myrzakhme, Alex P. James

Nazarbayev University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Citation (Scopus)

Abstract

Due to their flexible architecture and inherent parallelism, FPGAs are ideal candidates for neural network implementations. Still they have not achieved wide-spread acceptance in this regard. One of the major roadblocks for FPGAs is the implementation of complex mathematical functions encountered in neural networks. Exact implementation of these functions consume large number of resources. In this paper we discuss an FPGA-based neural network prototyping platform and the approximate implementation of a probabilistic neural network (PNN) on a Xilinx 7-Series FPGA. The complex mathematical functions as replaced by approximations. Analysis shows that hardware performance is much higher than that of software counter parts and the error induced due to approximations is within tolerable limit.

Original language	English
Title of host publication	Proceedings of the 2nd International Conference on Computing and Network Communications, CoCoNet 2018
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	64-68
Number of pages	5
ISBN (Electronic)	9781538659281
DOIs	https://doi.org/10.1109/CoCoNet.2018.8476889
Publication status	Published - Sep 28 2018
Event	2nd International Conference on Computing and Network Communications, CoCoNet 2018 - Astana, Kazakhstan Duration: Aug 15 2018 → Aug 17 2018

Conference

Conference	2nd International Conference on Computing and Network Communications, CoCoNet 2018
Country	Kazakhstan
City	Astana
Period	8/15/18 → 8/17/18

ASJC Scopus subject areas

Computer Networks and Communications
Computer Science Applications
Hardware and Architecture
Electrical and Electronic Engineering

Access to Document

10.1109/CoCoNet.2018.8476889

Fingerprint Dive into the research topics of 'FAPNN: An FPGA based Approximate Probabilistic Neural Network Library'. Together they form a unique fingerprint.

Cite this

FAPNN : An FPGA based Approximate Probabilistic Neural Network Library. / Vipin, Kizheppatt; Akhmetov, Yerbol; Myrzakhme, Serikbolsyn; James, Alex P.

Proceedings of the 2nd International Conference on Computing and Network Communications, CoCoNet 2018. Institute of Electrical and Electronics Engineers Inc., 2018. p. 64-68 8476889.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Vipin, K, Akhmetov, Y, Myrzakhme, S & James, AP 2018, FAPNN: An FPGA based Approximate Probabilistic Neural Network Library. in Proceedings of the 2nd International Conference on Computing and Network Communications, CoCoNet 2018., 8476889, Institute of Electrical and Electronics Engineers Inc., pp. 64-68, 2nd International Conference on Computing and Network Communications, CoCoNet 2018, Astana, Kazakhstan, 8/15/18. https://doi.org/10.1109/CoCoNet.2018.8476889

@inproceedings{c2f19333f3c741799ef01d837c2e8a77,

title = "FAPNN: An FPGA based Approximate Probabilistic Neural Network Library",

abstract = "Due to their flexible architecture and inherent parallelism, FPGAs are ideal candidates for neural network implementations. Still they have not achieved wide-spread acceptance in this regard. One of the major roadblocks for FPGAs is the implementation of complex mathematical functions encountered in neural networks. Exact implementation of these functions consume large number of resources. In this paper we discuss an FPGA-based neural network prototyping platform and the approximate implementation of a probabilistic neural network (PNN) on a Xilinx 7-Series FPGA. The complex mathematical functions as replaced by approximations. Analysis shows that hardware performance is much higher than that of software counter parts and the error induced due to approximations is within tolerable limit.",

author = "Kizheppatt Vipin and Yerbol Akhmetov and Serikbolsyn Myrzakhme and James, {Alex P.}",

year = "2018",

month = sep,

day = "28",

doi = "10.1109/CoCoNet.2018.8476889",

language = "English",

pages = "64--68",

booktitle = "Proceedings of the 2nd International Conference on Computing and Network Communications, CoCoNet 2018",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

address = "United States",

note = "2nd International Conference on Computing and Network Communications, CoCoNet 2018 ; Conference date: 15-08-2018 Through 17-08-2018",

}

TY - GEN

T1 - FAPNN

T2 - 2nd International Conference on Computing and Network Communications, CoCoNet 2018

AU - Vipin, Kizheppatt

AU - Akhmetov, Yerbol

AU - Myrzakhme, Serikbolsyn

AU - James, Alex P.

PY - 2018/9/28

Y1 - 2018/9/28

N2 - Due to their flexible architecture and inherent parallelism, FPGAs are ideal candidates for neural network implementations. Still they have not achieved wide-spread acceptance in this regard. One of the major roadblocks for FPGAs is the implementation of complex mathematical functions encountered in neural networks. Exact implementation of these functions consume large number of resources. In this paper we discuss an FPGA-based neural network prototyping platform and the approximate implementation of a probabilistic neural network (PNN) on a Xilinx 7-Series FPGA. The complex mathematical functions as replaced by approximations. Analysis shows that hardware performance is much higher than that of software counter parts and the error induced due to approximations is within tolerable limit.

AB - Due to their flexible architecture and inherent parallelism, FPGAs are ideal candidates for neural network implementations. Still they have not achieved wide-spread acceptance in this regard. One of the major roadblocks for FPGAs is the implementation of complex mathematical functions encountered in neural networks. Exact implementation of these functions consume large number of resources. In this paper we discuss an FPGA-based neural network prototyping platform and the approximate implementation of a probabilistic neural network (PNN) on a Xilinx 7-Series FPGA. The complex mathematical functions as replaced by approximations. Analysis shows that hardware performance is much higher than that of software counter parts and the error induced due to approximations is within tolerable limit.

UR - http://www.scopus.com/inward/record.url?scp=85055980488&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055980488&partnerID=8YFLogxK

U2 - 10.1109/CoCoNet.2018.8476889

DO - 10.1109/CoCoNet.2018.8476889

M3 - Conference contribution

AN - SCOPUS:85055980488

SP - 64

EP - 68

BT - Proceedings of the 2nd International Conference on Computing and Network Communications, CoCoNet 2018

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 15 August 2018 through 17 August 2018

ER -

FAPNN: An FPGA based Approximate Probabilistic Neural Network Library

Abstract

Conference

ASJC Scopus subject areas

Access to Document

Other files and links

Cite this