PaZoe: classifying time series with few labels

Kamalov, Mikhail; Boisbunon, Aurélie; Fanara, Carlo; Grenet, Ingrid; Daeden, Jonathan

doi:10.23919/eusipco54536.2021.9615924

Cited by 5 publications

(3 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These predictions are used as pseudo-labels of the unlabeled examples and used for retraining the classifier. A similar approach was presented in [7] where the authors combined the PageRank and PCA algorithms with a variant of genetic programming (GP) specifically tailored for non-linear symbolic regression. The algorithm was tested on three time series datasets and it was reported that the performance of the hybrid-algorithm overcomes the two algorithms individually.…”

Section: Neuroevolution For Semi-supervised Problemsmentioning

confidence: 99%

“…This metric is not expensive to compute since it only requires calculating the predictions of the model for all instances in D u train . The second baseline uses retraining, an approach frequently described in the literature for semisupervised learning [23,7]. Retraining consists of using the model learned on labeled instances to make predictions on the unlabeled instances.…”

Section: Baselines For Semi-supervised Classification With Neuroevolu...mentioning

confidence: 99%

“…The quality of the model is evaluated by combining the accuracy on the labeled data with one measure of cluster quality on the unlabeled examples. Kamalov et al [7] apply a GP algorithm combined with a self-labeling algorithm to classify time-series. They show that the combination of both types of algorithms improves state-of-the-art semi-supervised methods for the addressed classification problems.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Neuroevolutionary algorithms driven by neuron coverage metrics for semi-supervised classification

Santana¹,

Hidalgo-Cenalmor²,

Garciarena³

et al. 2023

Preprint

View full text Add to dashboard Cite

In some machine learning applications the availability of labeled instances for supervised classification is limited while unlabeled instances are abundant. Semisupervised learning algorithms deal with these scenarios and attempt to exploit the information contained in the unlabeled examples. In this paper, we address the question of how to evolve neural networks for semi-supervised problems. We introduce neuroevolutionary approaches that exploit unlabeled instances by using neuron coverage metrics computed on the neural network architecture encoded by each candidate solution. Neuron coverage metrics resemble code coverage metrics used to test software, but are oriented to quantify how the different neural network components are covered by test instances. In our neuroevolutionary approach, we define fitness functions that combine classification accuracy computed on labeled examples and neuron coverage metrics evaluated using unlabeled examples. We assess the impact of these functions on semi-supervised problems with a varying amount of labeled instances. Our results show that the use of neuron coverage metrics helps neuroevolution to become less sensitive to the scarcity of labeled data, and can lead in some cases to a more robust generalization of the learned classifiers.

show abstract

Section: Neuroevolution For Semi-supervised Problemsmentioning

confidence: 99%

Section: Baselines For Semi-supervised Classification With Neuroevolu...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation