-
A Unified Deep Learning Formalism For Processing Graph Signals
Authors:
Myriam Bontonou,
Carlos Lassance,
Jean-Charles Vialatte,
Vincent Gripon
Abstract:
Convolutional Neural Networks are very efficient at processing signals defined on a discrete Euclidean space (such as images). However, as they can not be used on signals defined on an arbitrary graph, other models have emerged, aiming to extend its properties. We propose to review some of the major deep learning models designed to exploit the underlying graph structure of signals. We express them…
▽ More
Convolutional Neural Networks are very efficient at processing signals defined on a discrete Euclidean space (such as images). However, as they can not be used on signals defined on an arbitrary graph, other models have emerged, aiming to extend its properties. We propose to review some of the major deep learning models designed to exploit the underlying graph structure of signals. We express them in a unified formalism, giving them a new and comparative reading.
△ Less
Submitted 1 May, 2019;
originally announced May 2019.
-
Matching Convolutional Neural Networks without Priors about Data
Authors:
Carlos Eduardo Rosar Kos Lassance,
Jean-Charles Vialatte,
Vincent Gripon
Abstract:
We propose an extension of Convolutional Neural Networks (CNNs) to graph-structured data, including strided convolutions and data augmentation on graphs.
Our method matches the accuracy of state-of-the-art CNNs when applied on images, without any prior about their 2D regular structure.
On fMRI data, we obtain a significant gain in accuracy compared with existing graph-based alternatives.
We propose an extension of Convolutional Neural Networks (CNNs) to graph-structured data, including strided convolutions and data augmentation on graphs.
Our method matches the accuracy of state-of-the-art CNNs when applied on images, without any prior about their 2D regular structure.
On fMRI data, we obtain a significant gain in accuracy compared with existing graph-based alternatives.
△ Less
Submitted 27 February, 2018;
originally announced February 2018.
-
Convolutional neural networks on irregular domains based on approximate vertex-domain translations
Authors:
Bastien Pasdeloup,
Vincent Gripon,
Jean-Charles Vialatte,
Dominique Pastor,
Pascal Frossard
Abstract:
We propose a generalization of convolutional neural networks (CNNs) to irregular domains, through the use of a translation operator on a graph structure. In regular settings such as images, convolutional layers are designed by translating a convolutional kernel over all pixels, thus enforcing translation equivariance. In the case of general graphs however, translation is not a well-defined operati…
▽ More
We propose a generalization of convolutional neural networks (CNNs) to irregular domains, through the use of a translation operator on a graph structure. In regular settings such as images, convolutional layers are designed by translating a convolutional kernel over all pixels, thus enforcing translation equivariance. In the case of general graphs however, translation is not a well-defined operation, which makes shifting a convolutional kernel not straightforward. In this article, we introduce a methodology to allow the design of convolutional layers that are adapted to signals evolving on irregular topologies, even in the absence of a natural translation. Using the designed layers, we build a CNN that we train using the initial set of signals. Contrary to other approaches that aim at extending CNNs to irregular domains, we incorporate the classical settings of CNNs for 2D signals as a particular case of our approach. Designing convolutional layers in the vertex domain directly implies weight sharing, which in other approaches is generally estimated a posteriori using heuristics.
△ Less
Submitted 5 November, 2018; v1 submitted 27 October, 2017;
originally announced October 2017.
-
A neighborhood-preserving translation operator on graphs
Authors:
Bastien Pasdeloup,
Vincent Gripon,
Jean-Charles Vialatte,
Nicolas Grelier,
Dominique Pastor
Abstract:
In this paper, we introduce translation operators on graphs. Contrary to spectrally-defined translations in the framework of graph signal processing, our operators mimic neighborhood-preserving properties of translation operators defined in Euclidean spaces directly in the vertex domain, and therefore do not deform a signal as it is translated. We show that in the case of grid graphs built on top…
▽ More
In this paper, we introduce translation operators on graphs. Contrary to spectrally-defined translations in the framework of graph signal processing, our operators mimic neighborhood-preserving properties of translation operators defined in Euclidean spaces directly in the vertex domain, and therefore do not deform a signal as it is translated. We show that in the case of grid graphs built on top of a metric space, these operators exactly match underlying Euclidean translations, suggesting that they completely leverage the underlying metric. More generally, these translations are defined on any graph, and can therefore be used to process signals on those graphs. We show that identifying proposed translations is in general an NP-Complete problem. To cope with this issue, we introduce relaxed versions of these operators, and illustrate translation of signals on random graphs.
△ Less
Submitted 3 December, 2018; v1 submitted 12 September, 2017;
originally announced September 2017.
-
Learning Local Receptive Fields and their Weight Sharing Scheme on Graphs
Authors:
Jean-Charles Vialatte,
Vincent Gripon,
Gilles Coppin
Abstract:
We propose a simple and generic layer formulation that extends the properties of convolutional layers to any domain that can be described by a graph. Namely, we use the support of its adjacency matrix to design learnable weight sharing filters able to exploit the underlying structure of signals in the same fashion as for images. The proposed formulation makes it possible to learn the weights of th…
▽ More
We propose a simple and generic layer formulation that extends the properties of convolutional layers to any domain that can be described by a graph. Namely, we use the support of its adjacency matrix to design learnable weight sharing filters able to exploit the underlying structure of signals in the same fashion as for images. The proposed formulation makes it possible to learn the weights of the filter as well as a scheme that controls how they are shared across the graph. We perform validation experiments with image datasets and show that these filters offer performances comparable with convolutional ones.
△ Less
Submitted 5 October, 2017; v1 submitted 8 June, 2017;
originally announced June 2017.
-
A Study of Deep Learning Robustness Against Computation Failures
Authors:
Jean-Charles Vialatte,
François Leduc-Primeau
Abstract:
For many types of integrated circuits, accepting larger failure rates in computations can be used to improve energy efficiency. We study the performance of faulty implementations of certain deep neural networks based on pessimistic and optimistic models of the effect of hardware faults. After identifying the impact of hyperparameters such as the number of layers on robustness, we study the ability…
▽ More
For many types of integrated circuits, accepting larger failure rates in computations can be used to improve energy efficiency. We study the performance of faulty implementations of certain deep neural networks based on pessimistic and optimistic models of the effect of hardware faults. After identifying the impact of hyperparameters such as the number of layers on robustness, we study the ability of the network to compensate for computational failures through an increase of the network size. We show that some networks can achieve equivalent performance under faulty implementations, and quantify the required increase in computational complexity.
△ Less
Submitted 18 April, 2017;
originally announced April 2017.
-
Neighborhood-Preserving Translations on Graphs
Authors:
Nicolas Grelier,
Bastien Pasdeloup,
Jean-Charles Vialatte,
Vincent Gripon
Abstract:
In many domains (e.g. Internet of Things, neuroimaging) signals are naturally supported on graphs. These graphs usually convey information on similarity between the values taken by the signal at the corresponding vertices. An interest of using graphs is that it allows to define ad hoc operators to perform signal processing. Among them, ones of paramount importance in many tasks are translations. I…
▽ More
In many domains (e.g. Internet of Things, neuroimaging) signals are naturally supported on graphs. These graphs usually convey information on similarity between the values taken by the signal at the corresponding vertices. An interest of using graphs is that it allows to define ad hoc operators to perform signal processing. Among them, ones of paramount importance in many tasks are translations. In this paper we are interested in defining translations on graphs using a few simple properties. Namely we propose to define translations as functions from vertices to adjacent ones, that preserve neighborhood properties of the graph. We show that our definitions, contrary to other works on the subject, match usual translations on grid graphs.
△ Less
Submitted 8 June, 2016;
originally announced June 2016.
-
Generalizing the Convolution Operator to extend CNNs to Irregular Domains
Authors:
Jean-Charles Vialatte,
Vincent Gripon,
Grégoire Mercier
Abstract:
Convolutional Neural Networks (CNNs) have become the state-of-the-art in supervised learning vision tasks. Their convolutional filters are of paramount importance for they allow to learn patterns while disregarding their locations in input images. When facing highly irregular domains, generalized convolutional operators based on an underlying graph structure have been proposed. However, these oper…
▽ More
Convolutional Neural Networks (CNNs) have become the state-of-the-art in supervised learning vision tasks. Their convolutional filters are of paramount importance for they allow to learn patterns while disregarding their locations in input images. When facing highly irregular domains, generalized convolutional operators based on an underlying graph structure have been proposed. However, these operators do not exactly match standard ones on grid graphs, and introduce unwanted additional invariance (e.g. with regards to rotations). We propose a novel approach to generalize CNNs to irregular domains using weight sharing and graph-based operators. Using experiments, we show that these models resemble CNNs on regular domains and offer better performance than multilayer perceptrons on distorded ones.
△ Less
Submitted 25 October, 2017; v1 submitted 3 June, 2016;
originally announced June 2016.
