Analysis of Negative Correlation Learning

This paper describes negative correlation learning for designing neural network ensembles. Negative correlation learning has been firstly analysed in terms of minimising mutual information on a regression task. By minimising the mutual information between variables extracted by two neural networks, they are forced to convey different information a-bout some features of their input. Based on the decision boundaries and correct response sets, negative correlation learning has been further studied on two pattern classification problems. The purpose of examining the decision boundaries and the correct response sets is not only to illustrate the learning behavior of negative correlation learning, but also to cast light on how to design more effective neural network ensembles. The experimental results showed the decision boundary of the trained neural network ensemble by negative correlation learning is almost as good as the optimum decision boundary.

A Heterogeneous Ensemble of Extreme Learning Machines with Correntropy and Negative Correlation

The Extreme Learning Machine（ELM） is an effective learning algorithm for a Single-Layer Feedforward Network（SLFN）. It performs well in managing some problems due to its fast learning speed. However, in practical applications, its performance might be affected by the noise in the training data. To tackle the noise issue, we propose a novel heterogeneous ensemble of ELMs in this article. Specifically, the correntropy is used to achieve insensitive performance to outliers, while implementing Negative Correlation Learning（NCL） to enhance diversity among the ensemble. The proposed Heterogeneous Ensemble of ELMs（HE2 LM） for classification has different ELM algorithms including the Regularized ELM（RELM）, the Kernel ELM（KELM）, and the L2-norm-optimized ELM（ELML2）. The ensemble is constructed by training a randomly selected ELM classifier on a subset of the training data selected through random resampling. Then, the class label of unseen data is predicted using a maximum weighted sum approach. After splitting the training data into subsets, the proposed HE2 LM is tested through classification and regression tasks on real-world benchmark datasets and synthetic datasets. Hence, the simulation results show that compared with other algorithms, our proposed method can achieve higher prediction accuracy, better generalization, and less sensitivity to outliers.

Ruin probability for correlated negative risk sums model with Erlang processes

This paper studies a Sparre Andersen negative risk sums model in which the distribution of ＂interclaim＂ time is that of a sum of n independent exponential random variables. Thus, the Erlang（n） model is a special case. On this basis the correlated negative risk sums process with the common Erlang process is considered. Integro-differential equations with boundary conditions for ψ（u） are given. For some special cases a closed-form expression for ψ（u） is derived.

Research on Feature Extraction Method of Social Network Text

The development of various applications based on social network text is in full swing.Studying text features and classifications is of great value to extract important information.This paper mainly introduces the common feature selection algorithms and feature representation methods,and introduces the basic principles,advantages and disadvantages of SVM and KNN,and the evaluation indexes of classification algorithms.In the aspect of mutual information feature selection function,it describes its processing flow,shortcomings and optimization improvements.In view of its weakness in not balancing the positive and negative correlation characteristics,a balance weight attribute factor and feature difference factor are introduced to make up for its deficiency.The experimental stage mainly describes the specific process:the word segmentation processing,to disuse words,using various feature selection algorithms,including optimized mutual information,and weighted with TF-IDF.Under the two classification algorithms of SVM and KNN,we compare the merits and demerits of all the feature selection algorithms according to the evaluation index.Experiments show that the optimized mutual information feature selection has good performance and is better than KNN under the SVM classification algorithm.This experiment proves its validity.

Precipitation Effects on Temperature—A Case Study in China

To discover the mechanism of the continuous summer temperature decline in certain re-gions against significant global warming,456 national meteorological observational stations with long term of observed daily temperature and precipitation data were applied to analyze the relationship between precipitation and temperature.Results show that there is a significant negative correlation between precipitation and temperature,indicating that precipitation influences temperature signifi-cantly：the more the precipitation,the lower the temperature,and vice versa.The precipitation effect on temperature can reasonably interpret the typical summer temperature in Southwest China and can partly explain why global warming is more significant in winter than in summer and more significant in higher latitude areas than in lower latitude.

Local proton hopping mechanism in imidazolium-based plastic crystal:an ab initio molecular dynamics study

Protic organic ionic plastic crystals(POIPCs)are promising solid-state proton conductor materials in anhydrous proton exchange membrane fuel cells,due to their mechanical flexibility and high ionic conductivity in the plastic crystal phase.In typical POIPCs,the ions are orientationally disordered while the centers of mass are ordered(positional order)like the crystal phase.The local disorder provides more degrees of freedom for the translational and rotational diffusion of ions,thus enhancing proton conduction either via the vehicle mechanism or the Grotthuss mechanism.Yet the local dynamics and the interactions of the cations and anions during the proton transfer process are far from being fully understood.Here,we performed Car–Parrinello molecular dynamics(CPMD)simulation on the imidazolium methanesulfate([ImH][CH_(3)SO_(3)])unit cell.By artificially creating one proton hole,we found that a proton can hop directly between the cations.Though the anion is not directly involved in proton hopping,the oxygen atom in the sulfonate group interacts with the proton and has a synergetic motion along with the proton hopping process.This indicates the structural disorder of imidazolium rings and the aid of an anion can facilitate Grotthuss-type proton hopping in imidazolium-based POIPCs.