Neural network method for determining embedding dimension of a time series

Maus, Adam; Sprott, Julien Clinton

doi:10.1016/j.cnsns.2010.10.030

Cited by 51 publications

(20 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…At 1,024 Hz, this provided a frequency resolution of 17 Hz, which was then zero-padded to a frequency resolution of 2 Hz. There are other methods to determine the embedding dimension (Cao, 1997; Maus and Sprott, 2011); it is beyond the scope of this paper to include an exhaustive discussion of these methods.…”

Section: Resultsmentioning

confidence: 99%

Multivariate cross-frequency coupling via generalized eigendecomposition

Cohen

2017

eLife

View full text Add to dashboard Cite

This paper presents a new framework for analyzing cross-frequency coupling in multichannel electrophysiological recordings. The generalized eigendecomposition-based cross-frequency coupling framework (gedCFC) is inspired by source-separation algorithms combined with dynamics of mesoscopic neurophysiological processes. It is unaffected by factors that confound traditional CFC methods—such as non-stationarities, non-sinusoidality, and non-uniform phase angle distributions—attractive properties considering that brain activity is neither stationary nor perfectly sinusoidal. The gedCFC framework opens new opportunities for conceptualizing CFC as network interactions with diverse spatial/topographical distributions. Five specific methods within the gedCFC framework are detailed, these are validated in simulated data and applied in several empirical datasets. gedCFC accurately recovers physiologically plausible CFC patterns embedded in noise that causes traditional CFC methods to perform poorly. The paper also demonstrates that spike-field coherence in multichannel local field potential data can be analyzed using the gedCFC framework, which provides significant advantages over traditional spike-field coherence analyses. Null-hypothesis testing is also discussed.DOI: http://dx.doi.org/10.7554/eLife.21792.001

show abstract

Section: Resultsmentioning

confidence: 99%

Multivariate cross-frequency coupling via generalized eigendecomposition

Cohen

2017

eLife

View full text Add to dashboard Cite

show abstract

“…It has been demonstrated in Maus and Sprott that E 1( m ) would remain almost stable when m is greater than the threshold m 0 (the minimal embedding dimension that we wish to obtain).…”

Section: Icing Prediction Modeling Using Psr and Multivariate Time Sementioning

confidence: 91%

“…There are many machine learning algorithms, eg, BPNN, self‐organizing map, RNN, RBM, and SVR, that can be used to solve this regression task . One of the most powerful nonlinear regression algorithms, SVR, is adopted in this paper to fit F τ .…”

Section: Icing Prediction Modeling Using Psr and Multivariate Time Sementioning

confidence: 99%

Online prediction method of icing of overhead power lines based on support vector regression

Miao

et al. 2018

Int Trans Electr Energ Syst

View full text Add to dashboard Cite

Summary This paper proposes a novel online multivariate time series prediction method, using support vector regression, to build an icing alert system that can forecast short‐term icing accretion load on overhead power lines. Conventional icing alert methods either cannot predict future icing accretion or these methods suffer from inadequate predictive accuracy. To resolve these issues, historical and online micrometeorological data from local observations are used to build a multivariate prediction model. Moreover, data preprocessing based on wavelets is used to prefilter spiking noise within the obtained field signals. In addition, the phase‐space reconstruction theory is applied to find the minimal embedding dimensions of the contributing factors by which the computational complexity of the multivariate model is reduced. Finally, an online adaptive predictive model based on support vector regression is proposed and implemented to further improve the predictive accuracy and predictive length of the icing process of overhead power lines. Experimental results indicate that this method can predict the real‐time icing value on overhead power lines 5 hours in advance, with an acceptable error term.

show abstract

“…The lag selection is performed as a postprocessing stage in Maus and Sprott (2011) with a sensitivity computation of the output to each time lag. The initial stage trains a single-layer, feed-forward ANN based on d time lags, with d chosen large enough to capture the relevant dynamics of the time series.…”

Section: Lags Selectionmentioning

confidence: 99%

Coevolution of lags and RBFNs for time series forecasting: L-Co-R algorithm

et al. 2011

View full text Add to dashboard Cite

This paper introduces Lags COevolving with Rbfns (L-Co-R), a coevolutionary method developed to face time-series forecasting problems. L-Co-R simultaneously evolves the model that provides the forecasted values and the set of time lags the model must use in the prediction process. Coevolution takes place by means of two populations that evolve at the same time, cooperating between them; the first population is composed of radial basis function neural networks; the second one contains the individuals representing the sets of lags. Thus, the final solution provided by the method comprises both the neural net and the set of lags that better approximate the time series. The method has been tested across 34 different time series datasets, and the results compared to 6 different methods referenced in literature, and with respect to 4 different error measures. The results show that L-Co-R outperforms the rest of methods, as the statistical analysis carried out indicates.

show abstract

Neural network method for determining embedding dimension of a time series

Cited by 51 publications

References 16 publications

Multivariate cross-frequency coupling via generalized eigendecomposition

Multivariate cross-frequency coupling via generalized eigendecomposition

Online prediction method of icing of overhead power lines based on support vector regression

Coevolution of lags and RBFNs for time series forecasting: L-Co-R algorithm

Contact Info

Product

Resources

About