Imbalanced evolving self-organizing learning

Cai, Qiao; He, Haibo; Man, Hong

doi:10.1016/j.neucom.2013.11.010

Cited by 35 publications

(25 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Rather than combining two different optimization methods, most of the researchers modified the original optimization method to obtain better results in the analysis [18][19][20][21]. T. Niknam [22] modified the honeybee optimization method in solving the multiobjective problem for DG allocation and sizing.…”

Section: Related Researchmentioning

confidence: 99%

Optimal multiple distributed generation output through rank evolutionary particle swarm optimization

2015

View full text Add to dashboard Cite

a b s t r a c tThe total power losses in a distribution network are usually minimized through the adjustment of the output of a distributed generator (DG). In line with this objective, most researchers concentrate on the optimization technique in order to regulate the DG 0 s output and compute its optimal size. In this article, a novel Rank Evolutionary Particle Swarm Optimization (REPSO) method is introduced. By hybridizing the Evolutionary Programming (EP) in Particle Swarm Optimization (PSO) algorithm, it will allow the entire particles to move toward the optimal value faster than usual and reach the convergence value. Moreover, the local best (P best ) and global best (G best ) values are obtained in simplify manner in the REPSO algorithm. The performance of this new algorithm will be compared to 3 well-known PSO methods, which are Conventional Particle Swarm Optimization (CPSO), Inertia Weight Particle Swarm Optimization (IWPSO), and Iteration Particle Swarm Optimization (IPSO) on 10 mathematical benchmark functions, and solving the optimal DG output problem. From the results, the IWPSO, IPSO and REPSO methods gave the similar "best" value in all functions after being tested 50 times, except for Function 6. However, the REPSO algorithm provided the lowest SD value in all problems. In the power system analysis, the performance of REPSO is similar to IWPSO and IPSO, and better than CPSO, but the REPSO algorithm requires less numbers of iteration and computing time. It can be concluded that the REPSO is a superior method in solving low dimension analysis, either in numerical optimization problems, or DG sizing problems.

show abstract

Section: Related Researchmentioning

confidence: 99%

Optimal multiple distributed generation output through rank evolutionary particle swarm optimization

2015

View full text Add to dashboard Cite

show abstract

“…Table 1 gathers only dynamic features used in the proposed investigation. It should be noticed that, in practice, at least 40 dynamic features can be x-velocity f 8 Measurement time f 4 y-velocity f 9 Pen-up time f 5 x, y -velocity f 10 Pen-down time recorded by a tablet or additionally calculated [20]. For example, acceleration in both the x and y directions can be computed on the basis of velocity and time measures.…”

Section: The Dynamic Signature Featuresmentioning

confidence: 99%

“…To accomplish this task the well known Dynamic Time Warping (DTW) technique was Table 2: List of similarity measures or coefficients [20] Coeff./measure Coefficient or similarity measure name Coeff./measure Coefficient or similarity measure name ω 1 Euclidean ω 10 Jaccard ω 2 Gower ω 11 Fidelity ω 3 Minkowski ω 12 Bhattacharyya ω 4 City Block ω 13 Hellinger ω 5 Cosine ω 14 Matusita ω 6 Kulczynski ω 15 Pearsonχ 2 ω 7…”

Section: Similarity Coefficientsmentioning

confidence: 99%

Dynamic signature verification method based on association of features with similarity measures

2016

View full text Add to dashboard Cite

“…Cai et al [9] proposed a hybrid learning model using a modified selforganizing maps algorithm. This method assigns a winner neuron based on an energy function minimizing local error in the competitive learning phase.…”

Section: Sentiment and Emotion Classificationmentioning

confidence: 99%

Word Embedding Composition for Data Imbalances in Sentiment and Emotion Classification

Chen

Xia

et al. 2015

Cogn Comput

View full text Add to dashboard Cite

Text classification often faces the problem of imbalanced training data. This is true in sentiment analysis and particularly prominent in emotion classification where multiple emotion categories are very likely to produce naturally skewed training data. Different sampling methods have been proposed to improve classification performance by reducing the imbalance ratio between training classes. However, data sparseness and the small disjunct problem remain obstacles in generating new samples for minority classes when the data are skewed and limited. Methods to produce meaningful samples for smaller classes rather than simple duplication are essential in overcoming this problem. In this paper, we present an oversampling method based on word embedding compositionality which produces meaningful balanced training data. We first use a large corpus to train a continuous skip-gram model to form a word embedding model maintaining the syntactic and semantic integrity of the word features. Then, a compositional algorithm based on recursive neural tensor networks is used to construct sentence vectors based on the word embedding model. Finally, we use the SMOTE algorithm as an oversampling method to generate samples for the minority classes and produce a fully balanced training set. Evaluation results on two quite different tasks show that the feature composition method and the oversampling method are both important in obtaining improved classification results. Our method effectively addresses the data imbalance issue and consequently achieves improved results for both sentiment and emotion classification.

show abstract

Imbalanced evolving self-organizing learning

Cited by 35 publications

References 36 publications

Optimal multiple distributed generation output through rank evolutionary particle swarm optimization

Optimal multiple distributed generation output through rank evolutionary particle swarm optimization

Dynamic signature verification method based on association of features with similarity measures

Word Embedding Composition for Data Imbalances in Sentiment and Emotion Classification

Contact Info

Product

Resources

About