China
Africa
Explore chapters and articles related to this topic
Literature review and proposed framework
Published in Juan Carlos Chacon-Hurtado, Optimisation of Dynamic Heterogeneous Rainfall Sensor Networks in the Context of Citizen Observatories, 2019
Where I(X1, X2, …, Xn) is the multivariate mutual information among n variables, and I(X1, X2, …, Xn-1 | Xn) is the conditional information of n-1 variables with respect to the nth variable. The conditional mutual information can be understood as the amount of information that a set of variable share with another variable (or variables). The conditional mutual information of two variables (X1 and X2) with respect to a third one (X3) can be quantified as: I(X1,X2|X3)=H(X1|X3)−H(X1|X2,X3)
Prediction of dissolved gas concentration in power transformer oil based on the multivariate ReLu deep belief networks model
Published in Rodolfo Dufo-López, Jaroslaw Krzywanski, Jai Singh, Emerging Developments in the Power and Energy Industry, 2019
Lei Su, Lu Chen, Peng Xu, Yingjie Yan, Gehao Sheng
Many scholars in China and abroad have performed a great deal of research on the prediction of dissolved gas concentration in oil. Generally, the traditional methods are gray model (GM), artificial neural network (ANN), support vector machine (SVM), and other methods (Luo 2001; Yang 2008; Bian 2012; Fu 2013). The traditional method often only considers the development trend of a certain gas in modeling, ignoring the correlation analysis between gases, making the prediction effect not scientific enough and the stability is poor. In order to solve this problem, Xiao (2006), Sima (2012), and Lin (2016) introduced the gray relational calculation, taking the correlation between the components into account, using the gray correlation analysis to predict the correlation of the input variables before the prediction, and eliminating the weaker correlation factors. Then the above prediction method was used to model the calculation. In addition, Tang (2013) improved the standardized mutual information variable selection algorithm for the selection of input variables, and introduced conditional mutual information to replace the original mutual information items in the feature selection evaluation standard, which improved the selection validity of input variables to some extent. Although the gray correlation and mutual information technology avoids the defects of single-component gas concentration prediction, the calculation amount is increased in the data preprocessing process, and the correlation threshold has a certain subjectivity. With the development of deep learning methods, the feature extraction ability of a deep belief network (DBN) shows certain characteristics in transformer fault classification modeling, pattern recognition, time series prediction, etc. Shi et al. (2016) provide new ideas for data processing in the power sector.
Hybrid Approach towards Malaria Parasites Detection from Thin Blood Smear Image
Published in Siddhartha Bhattacharyya, Anirban Mukherjee, Indrajit Pan, Paramartha Dutta, Arup Kumar Bhaumik, Hybrid Intelligent Techniques for Pattern Analysis and Understanding, 2017
Sanjay Nag, Nabanita Basu, Samir Kumar Bandyopadhyay
The conditional mutual information maximization algorithm was used to find a filter feature set that bring in the maximum information about the concerned class label. This also encompasses or takes into account the nonlinear relationship between a feature vector and the class label.
A data-driven robust optimization method for the assembly job-shop scheduling problem under uncertainty
Published in International Journal of Computer Integrated Manufacturing, 2022
Peng Zheng, Peng Zhang, Junliang Wang, Jie Zhang, Changqi Yang, Yongqiao Jin
Given an AJSSP instance with P products and m machines, the assembly structure of each product and the standard processing time of each operation are known. First of all, a set of feasible schedules (not necessarily optimal or near-optimal) for the instances with the same number of operations are randomly generated as training samples. Let denote the schedule set and let denote the number of all schedules of . For each schedule , it is represented by a feature vector , which contains K features that can reflect the fluctuation law of makespan. These K key features are selected in a candidate feature set (Mirshekarian and Šormaz 2016) by a conditional mutual information-based feature selection method that we proposed in Wang and Zhang (2016).
Big data driven cycle time parallel prediction for production planning in wafer manufacturing
Published in Enterprise Information Systems, 2018
Junliang Wang, Jungang Yang, Jie Zhang, Xiaoxi Wang, Wenjun (Chris) Zhang
With the preprocessed data, a conditional mutual information based feature selection method (Wang and Zhang 2016) is performed to determine the key factors as the input factors of the prediction model. As a result, 78 factors with high correlation are obtained (shown in Table 5) in five categories to be the input factors in the numerical experiments, such as the queuing time before some photolithography machines, the number of work in process, the priority of each wafer lot.