170,078 research outputs found
Spectral clustering with graph neural networks for graph pooling
No full textSpectral clustering (SC) is a popular clustering
technique to find strongly connected communities on a graph. SC can be used in Graph Neural Networks (GNNs) to implement pooling operations that aggregate nodes belonging to the
same cluster. However, the eigendecomposition
of the Laplacian is expensive and, since clustering results are graph-specific, pooling methods
based on SC must perform a new optimization
for each new sample. In this paper, we propose
a graph clustering approach that addresses these
limitations of SC. We formulate a continuous relaxation of the normalized minCUT problem and
train a GNN to compute cluster assignments that
minimize this objective. Our GNN-based implementation is differentiable, does not require to
compute the spectral decomposition, and learns
a clustering function that can be quickly evaluated on out-of-sample graphs. From the proposed
clustering method, we design a graph pooling operator that overcomes some important limitations
of state-of-the-art graph pooling techniques and
achieves the best performance in several supervised and unsupervised tasks
Cluster partitioning in image analysis classification: A genetic algorithm approach
No full textA classification of data by using the genetic algorithm computational paradigm is proposed. The best data partition is defined to be the one minimizing the sum of Pythagorean distances between each datum in a cluster and the relative center of class or center of mass. Background is given, and the relevant genetic algorithm description is provided. The model for the genetic application is presented. Simulation results confirm genetic algorithms to be powerful tools for the solution of optimization problems
A 'combination of multiple classifier' design for low-complex, highly performing and power-aware classifiers
No full textIn this paper we study the relationships among the Combination of Multiple Classifier design philosophy, application level properties such as temporal and spatial locality of the inputs and low level aspects immediately impacting on power consumption, cache miss and computational complexity reduction. The CMC structure requires a set of independent simple sub-classifiers, each of which ruling in an application sub-domain under the control of a master enabling module and is particularly appealing in embedded system implementation. Only a sub-classifier is active at a time, the others being switched off
Guest Editorial - Special issue on neural technologies for identification, control, robotics, and signal/image processing
No full textGuest editorial - Special issue on neural technologies for identification, control, robotics, and signal/image processing
No full textA Neural-Network Based Control Solution to Air-Fuel Ratio Control for Automotive Fuel-Injection Systems
No full textMaximization of the catalyst efficiency in automotive fuel-injection engines requires the design of accurate control systems to keep the air-to-fuel ratio at the optimal stoichiometric value AFS. Unfortunately, this task is complex since the air-to-fuel ratio is very sensitive to small perturbations of the engine parameters. Some mechanisms ruling the engine and the combustion process are in fact unknown and/or show hard nonlinearities. These difficulties limit the effectiveness of traditional control approaches. In this paper, we suggest a neural based solution to the air-to-fuel ratio control in fuel injection systems. An indirect control approach has been considered which requires a preliminary modeling of the engine dynamics. The model for the engine and the final controller are based on recurrent neural networks with external feedbacks. Requirements for feasible control actions and the static precision of control have been integrated in the controller design to guide learning toward an effective control solution
Neural modeling of dynamic systems with non-measurable state variables
No full textThe paper studies the ability possessed by recurrent neural networks to model dynamic systems when some relevant state variables are not measurable. Neural architectures based on virtual states - which naturally arise from a space state representation - are introduced and compared with the more traditional neural output error ones. Despite the evident potential model ability possessed by virtual state architectures we experimented that their performances strongly depend on the training efficiency. A novel validation criterion for neural output error architectures is suggested which allows to assess the neural network not only in terms of its approximation accuracy but also with respect to stability issues
Experimental neural networks for prediction and identification
No full textIn this paper we prove the effectiveness of using simple NARX-type (nonlinear auto-regressive model with exogenous variables) recurrent neural networks to identify time series and nonlinear dynamical systems. Experimentally we show that, whenever the process generating the data is ruled by a linear model, the performances provided by the neural network are comparable with the ones given by the optimal predictor determined according to the Kolmogorov-Wiener theory. On the other hand, whenever the system to be modelled is intrinsically nonlinear, its performance approaches that obtainable with classical linear identification. The work extends that suggested by Narendra in (1990) by considering a reduced set of training data and a black-box model for the system to be identified
Graph Neural Networks in TensorFlow and Keras with Spektral
No full textGraph neural networks have enabled the application of deep learning to problems that can be described by graphs, which are found throughout the different fields of science, from physics to biology, natural language processing, telecommunications or medicine. In this paper we present Spektral, an open-source Python library for building graph neural networks with TensorFlow and the Keras application programming interface. Spektral implements a large set of methods for deep learning on graphs, including message-passing and pooling operators, as well as utilities for processing graphs and loading popular benchmark datasets. The purpose of this library is to provide the essential building blocks for creating graph neural networks, focusing on the guiding principles of user-fr iendliness and quick prototyping on which Keras is based. Spektral is, therefore, suitable for absolute beginners and expert deep learning practitioners alike. In this work, we present an overview of Spektral’s features and report the performance of the methods implemented by the library in scenarios of node classification, graph classification, and graph regression
- …
