Study of Learning Entropy for onset detection of epileptic seizures in EEG time series

Bukovský, Ivo; Cejnek, Matouš; Vrba, Jan; Homma, Noriyasu

doi:10.1109/ijcnn.2016.7727621

Cited by 6 publications

(6 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Later on, cumulative sums were proposed for practical approximations of LE, ie,

LE (k) = \frac{1}{n_{α} \cdot n_{w}} \sum_{\forall α} \sum_{i = 1}^{n_{w}} Γ (| △ w_{i} | > α \cdot \overset{true‾}{| △ w_{i} |}),

where

Γ = \{\begin{matrix} 1, & if false| △ {boldw}_{i} false| > α \cdot \overset{false| △ {boldw}_{i} false|}{true} \\ 0, & otherwise \end{matrix}

is an indicator function, and

\overset{false| △ {boldw}_{i} false|}{true}

is the mean absolute value of the recent history of the i th weight adaptation increment (for details, see Section 3) and α is the multiscale sensitivity parameter α ∈ { α 1 , α 2 ,…, α n }, α 1 < α 2 < ⋯ < α n . Figure recalls an intriguing performance of LE performed with significant robustness against false detection and with much lower computational complexity than the sample entropy …”

Section: Learning Entropymentioning

confidence: 72%

“…Comparison of Learning Entropy (LE) vs Sample Entropy (SE) for novelty detection in electroencephalogram (EEG) signal with an epileptic seizure; LE performs more robust seizure onset detection of a seizure with a lower level of computational complexity (adopted and modified from the work of Bukovsky et al)…”

Section: Learning Entropymentioning

confidence: 99%

“…Another fundamental approach is Sample Entropy algorithm . A comparison between these two methods can be found in the work of Bukovsky et al However, the first algorithm to estimate LE was proposed with respect to unusually extensive learning effort induced due to novelty in the data, while an unusually small decrease in the learning effort, such as if the noise suddenly disappearing from a deterministic signal, was not considered. In this paper, we introduce sequential detection methods well known from multivariate SPC, enabling us to enhance LE estimation and to improve LE‐based novelty detection.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Novelty detection based on learning entropy

Dohnal

Bukovský

2019

Appl Stoch Models Bus & Ind

View full text Add to dashboard Cite

The Approximate Individual Sample Learning Entropy is based on incremental learning of a predictor truex˜false(k+hfalse)=ϕfalse(boldxfalse(kfalse),boldwfalse), where x(k) is an input vector of a given size at time k, w is a vector of weights (adaptive parameters), and h is a prediction horizon. The basic assumption is that, after the underlying process x changes its behavior, the incrementally learning system will adapt the weights w to improve the predictor truex˜. Our goal is to detect a change in the behavior of the weight increment process. The main idea of this paper is based on the fact that weight increments △w(k), where △w(k) = w(k + 1) − w(k), create a weakly stationary process until a change occurs. Once a novelty behavior of the underlying process x(k) occurs, the process △w(k) changes its characteristics (eg, the mean or variation). We suggest using convenient characteristics of △w(k) in a multivariate detection scheme (eg, the Hotelling's T2 control chart).

show abstract

“…Later on, cumulative sums were proposed for practical approximations of LE, ie,

LE (k) = \frac{1}{n_{α} \cdot n_{w}} \sum_{\forall α} \sum_{i = 1}^{n_{w}} Γ (| △ w_{i} | > α \cdot \overset{true‾}{| △ w_{i} |}),

where

Γ = \{\begin{matrix} 1, & if false| △ {boldw}_{i} false| > α \cdot \overset{false| △ {boldw}_{i} false|}{true} \\ 0, & otherwise \end{matrix}

is an indicator function, and

\overset{false| △ {boldw}_{i} false|}{true}

Section: Learning Entropymentioning

confidence: 72%

Section: Learning Entropymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Novelty detection based on learning entropy

Dohnal

Bukovský

2019

Appl Stoch Models Bus & Ind

View full text Add to dashboard Cite

show abstract

“…Novelty detection via LE is based on real-time learning of systems after they had been pretrained on an initial pretraining data set. Aside from the founding work [34], other examples of works that indicate the usefulness of LE in biomedical or technical data and for novelty detection in data with concept drift can be found in [36][37][38][39].…”

Section: Introductionmentioning

confidence: 99%

Learning Entropy as a Learning-Based Information Concept

Bukovský

Kinsner

Homma

2019

Entropy

Self Cite

View full text Add to dashboard Cite

Recently, a novel concept of a non-probabilistic novelty detection measure, based on a multi-scale quantification of unusually large learning efforts of machine learning systems, was introduced as learning entropy (LE). The key finding with LE is that the learning effort of learning systems is quantifiable as a novelty measure for each individually observed data point of otherwise complex dynamic systems, while the model accuracy is not a necessary requirement for novelty detection. This brief paper extends the explanation of LE from the point of an informatics approach towards a cognitive (learning-based) information measure emphasizing the distinction from Shannon’s concept of probabilistic information. Fundamental derivations of learning entropy and of its practical estimations are recalled and further extended. The potentials, limitations, and, thus, the current challenges of LE are discussed.

show abstract

“…The Learning Entropy reflects the effort of the adaptive system to learn the novelty in the data. Authors in their later works extend original algorithm with prediction error evaluation [5] or with the different technique of Learning Entropy estimation [6] and shows, that evaluation of weight increments can be a useful approach to novelty detection topic [7], [8].…”

Section: Introductionmentioning

confidence: 99%

Adaptive Novelty Detection with Generalized Extreme Value Distribution

Vrba

2018

2018 International Conference on Applied Electronics (AE)

Self Cite

View full text Add to dashboard Cite

This paper introduces the new adaptive novelty detection method. The proposed method is using generalized extreme value distribution to evaluate the absolute value of adaptive system weight increments in time. The detection of novelty is threshold-based and the threshold ζ corresponds to the value of joint probability density function. Performance of the proposed algorithm is shown on artificial data. For comparison also results of Learning Entropy algorithm are shown, as this algorithm also evaluates the increments of adaptive weights.

show abstract

Study of Learning Entropy for onset detection of epileptic seizures in EEG time series

Cited by 6 publications

References 15 publications

Novelty detection based on learning entropy

Novelty detection based on learning entropy

Learning Entropy as a Learning-Based Information Concept

Adaptive Novelty Detection with Generalized Extreme Value Distribution

Contact Info

Product

Resources

About