PAC-Bayesian Margin Bounds for Convolutional Neural Networks - Technical Report

Pitas Konstantinos; Mike Davies; Pierre Vandergheynst

PAC-Bayesian Margin Bounds for Convolutional Neural Networks - Technical Report

Pitas Konstantinos, Mike Davies, Pierre Vandergheynst

School of Engineering

Research output: Working paper

Abstract / Description of output

Recently the generalisation error of deep neural networks has been analysed through the PAC-Bayesian framework, for the case of fully connected layers. We adapt this approach to the convolutional setting.

Original language	Undefined/Unknown
Publication status	Published - 30 Dec 2017

Keywords / Materials (for Non-textual outputs)

cs.LG
stat.ML

Access to Document

https://arxiv.org/pdf/1801.00171.pdf

C-SENSE: Exploiting low dimensional models in sensing, computation and signal processing
Davies, M.
EU government bodies
1/09/16 → 31/08/22
Project: Research
SpaRTaN: SpaRTaN: Sparse Representations and Compressed Sensing Training Network
Davies, M.
EU government bodies
1/10/14 → 30/09/18
Project: Research

Cite this

@techreport{7cc90fd57a524528b7c3a821edd7c859,

title = "PAC-Bayesian Margin Bounds for Convolutional Neural Networks - Technical Report",

abstract = "Recently the generalisation error of deep neural networks has been analysed through the PAC-Bayesian framework, for the case of fully connected layers. We adapt this approach to the convolutional setting.",

keywords = "cs.LG, stat.ML",

author = "Pitas Konstantinos and Mike Davies and Pierre Vandergheynst",

year = "2017",

month = dec,

day = "30",

language = "Undefined/Unknown",

type = "WorkingPaper",

}

TY - UNPB

T1 - PAC-Bayesian Margin Bounds for Convolutional Neural Networks - Technical Report

AU - Konstantinos, Pitas

AU - Davies, Mike

AU - Vandergheynst, Pierre

PY - 2017/12/30

Y1 - 2017/12/30

N2 - Recently the generalisation error of deep neural networks has been analysed through the PAC-Bayesian framework, for the case of fully connected layers. We adapt this approach to the convolutional setting.

AB - Recently the generalisation error of deep neural networks has been analysed through the PAC-Bayesian framework, for the case of fully connected layers. We adapt this approach to the convolutional setting.

KW - cs.LG

KW - stat.ML

M3 - Working paper

BT - PAC-Bayesian Margin Bounds for Convolutional Neural Networks - Technical Report

ER -

PAC-Bayesian Margin Bounds for Convolutional Neural Networks - Technical Report

Abstract / Description of output

Keywords / Materials (for Non-textual outputs)

Access to Document

Projects

C-SENSE: Exploiting low dimensional models in sensing, computation and signal processing

SpaRTaN: SpaRTaN: Sparse Representations and Compressed Sensing Training Network

Cite this