Search results for "machine learning."
showing 10 items of 1455 documents
Neural networks with non-uniform embedding and explicit validation phase to assess Granger causality
2015
A challenging problem when studying a dynamical system is to find the interdependencies among its individual components. Several algorithms have been proposed to detect directed dynamical influences between time series. Two of the most used approaches are a model-free one (transfer entropy) and a model-based one (Granger causality). Several pitfalls are related to the presence or absence of assumptions in modeling the relevant features of the data. We tried to overcome those pitfalls using a neural network approach in which a model is built without any a priori assumptions. In this sense this method can be seen as a bridge between model-free and model-based approaches. The experiments perfo…
Active spike transmission in the neuron model with a winding threshold manifold
2012
International audience; We analyze spiking responses of excitable neuron model with a winding threshold manifold on a pulse stimulation. The model is stimulated with external pulse stimuli and can generate nonlinear integrate-and-fire and resonant responses typical for excitable neuronal cells (all-or-none). In addition we show that for certain parameter range there is a possibility to trigger a spiking sequence with a finite number of spikes (a spiking message) in the response on a short stimulus pulse. So active transformation of N incoming pulses to M (with M>N) outgoing spikes is possible. At the level of single neuron computations such property can provide an active "spike source" comp…
Contextual neural-network based spectrum prediction for cognitive radio
2015
Cognitive radio is the technique of effective electromagnetic spectrum usage important for future wireless communication including 5G networks. Neural networks are nature-inspired computational models used to solve cognitive radio prediction problems. This paper presents the use of contextual Sigma-if neural network in prediction of channel states for cognitive radio. Our results indicate that Sigma-if neural network confirms better predictions than Multilayer Perceptron (MLP) network and decreases sensing time for the benefit of the increase of the effectiveness of e-m spectrum usage.
Incremental Generalized Discriminative Common Vectors for Image Classification.
2015
Subspace-based methods have become popular due to their ability to appropriately represent complex data in such a way that both dimensionality is reduced and discriminativeness is enhanced. Several recent works have concentrated on the discriminative common vector (DCV) method and other closely related algorithms also based on the concept of null space. In this paper, we present a generalized incremental formulation of the DCV methods, which allows the update of a given model by considering the addition of new examples even from unseen classes. Having efficient incremental formulations of well-behaved batch algorithms allows us to conveniently adapt previously trained classifiers without th…
On the Online Classification of Data Streams Using Weak Estimators
2016
In this paper, we propose a novel online classifier for complex data streams which are generated from non-stationary stochastic properties. Instead of using a single training model and counters to keep important data statistics, the introduced online classifier scheme provides a real-time self-adjusting learning model. The learning model utilizes the multiplication-based update algorithm of the Stochastic Learning Weak Estimator (SLWE) at each time instant as a new labeled instance arrives. In this way, the data statistics are updated every time a new element is inserted, without requiring that we have to rebuild its model when changes occur in the data distributions. Finally, and most impo…
Day-ahead forecasting for photovoltaic power using artificial neural networks ensembles
2016
Solar photovoltaic plants power output forecasting using machine learning techniques can be of a great advantage to energy producers when they are implemented with day-ahead energy market data. In this work a model was developed using a supervised learning algorithm of multilayer perceptron feedforward artificial neural network to predict the next twenty-four hours (day-ahead) power of a solar facility using fetched weather forecast of the following day. Each set of tested network configuration was trained by the historical power output of the plant as a target. For each configuration, one hundred networks ensembles was averaged to give the ability to generalize a better forecast. The train…
Towards more Valid Assessment of Learning from Animations
2020
Animated explanations have become an ubiquitous feature of modern educational practice. They provide a distinctive, non-verbal means of presenting information that is particularly appropriate for dynamic subject matter. However, the prevailing approaches used to assess learning from educational animations are almost exclusively verbal. There is thus a clear disconnect between the form of representation students encounter during their learning activity and the very different form of representation used to assess the resulting learning outcomes. This fundamental inconsistency undermines the validity of current assessment approaches and signals the need for a fresh look at how learning from an…
Local operators to detect regions of interest
1997
The performance of a visual system is strongly influenced by the information processing that is done in the early vision phase. The need exists to limit the computation on areas of interest to reduce the total amount of data and their redundancy. This paper describes a new method to drive the attention during the analysis of complex scenes. Two new local operators, based on the computation of local moments and symmetries, are combined to drive the selection. Experimental results on real data are also reported. © 1997 Elsevier Science B.V.
Approximation of functions over manifolds : A Moving Least-Squares approach
2021
We present an algorithm for approximating a function defined over a $d$-dimensional manifold utilizing only noisy function values at locations sampled from the manifold with noise. To produce the approximation we do not require any knowledge regarding the manifold other than its dimension $d$. We use the Manifold Moving Least-Squares approach of (Sober and Levin 2016) to reconstruct the atlas of charts and the approximation is built on-top of those charts. The resulting approximant is shown to be a function defined over a neighborhood of a manifold, approximating the originally sampled manifold. In other words, given a new point, located near the manifold, the approximation can be evaluated…
Optimal rates of convergence for persistence diagrams in Topological Data Analysis
2013
Computational topology has recently known an important development toward data analysis, giving birth to the field of topological data analysis. Topological persistence, or persistent homology, appears as a fundamental tool in this field. In this paper, we study topological persistence in general metric spaces, with a statistical approach. We show that the use of persistent homology can be naturally considered in general statistical frameworks and persistence diagrams can be used as statistics with interesting convergence properties. Some numerical experiments are performed in various contexts to illustrate our results.