Discriminant Rule Research Articles

High dimensional discriminant rules with shrinkage estimators of the covariance matrix and mean vector

Linear discriminant analysis (LDA) is a typical method for classification problems with large dimensions and small samples. There are various types of LDA methods that are based on the different types of estimators for the covariance matrices and mean vectors. In this paper, we consider shrinkage methods based on a non-parametric approach. For the precision matrix, methods based on the sparsity structure or data splitting are examined. Regarding the estimation of mean vectors, Non-parametric Empirical Bayes (NPEB) methods and Non-parametric Maximum Likelihood Estimation (NPMLE) methods, also known as f-modeling and g-modeling, respectively, are adopted. The performance of linear discriminant rules based on combined estimation strategies of the covariance matrix and mean vectors are analyzed in this study. Particularly, the study presents a theoretical result on the performance of the NPEB method and compares it with previous studies. Simulation studies with various covariance matrices and mean vector structures are conducted to evaluate the methods discussed in this paper. Furthermore, real data examples such as gene expressions and EEG data are also presented.

Learning-induced bidirectional enhancement of inhibitory synaptic metaplasticity.

Training rodents in a particularly difficult olfactory-discrimination (OD) task results in the acquisition of the ability to perform the task well, termed 'rule learning'. In addition to enhanced intrinsic excitability and synaptic excitation in piriform cortex pyramidal neurons, rule learning results in increased synaptic inhibition across the whole cortical network to the point where it precisely maintains the balance between inhibition and excitation. The mechanism underlying such precise inhibitory enhancement remains to be explored. Here, we use brain slices from transgenic mice (VGAT-ChR2-EYFP), enabling optogenetic stimulation of single GABAergic neurons and recordings of unitary synaptic events in pyramidal neurons. Quantal analysis revealed that learning-induced enhanced inhibition is mediated by increased quantal size of the evoked inhibitory events. Next, we examined the plasticity of synaptic inhibition induced by long-lasting, intrinsically evoked spike firing in post-synaptic neurons. Repetitive depolarizing current pulses from depolarized (-70mV) or hyperpolarized (-90mV) membrane potentials induced long-term depression (LTD) and long-term potentiation (LTP) of synaptic inhibition, respectively. We found a profound bidirectional increase in the ability to induce both LTD, mediated by L-type calcium channels, and LTP, mediated by R-type calcium channels after rule learning. Blocking the GABAB receptor reversed the effect of intrinsic stimulation at -90mV from LTP to LTD. We suggest that learning greatly enhances the ability to modify the strength of synaptic inhibition of principal neurons in both directions. Such plasticity of synaptic plasticity allows fine-tuning of inhibition on each particular neuron, thereby stabilizing the network while maintaining the memory of the rule. KEY POINTS: Olfactory discrimination rule learning results in long-lasting enhancement of synaptic inhibition on piriform cortex pyramidal neurons. Quantal analysis of unitary inhibitory synaptic events, evoked by optogenetic minimal stimulation, revealed that enhanced synaptic inhibition is mediated by increased quantal size. Surprisingly, metaplasticity of synaptic inhibition, induced by intrinsically evoked repetitive spike firing, is increased bidirectionally. The susceptibility to both long-term depression (LTD) and long-term potentiation (LTP) of inhibition is enhanced after learning. LTD of synaptic inhibition is mediated by L-type calcium channels and LTP by R-type calcium channels. LTP is also dependent on activation of GABAB receptors. We suggest that learning-induced changes in the metaplasticity of synaptic inhibition enable the fine-tuning of inhibition on each particular neuron, thereby stabilizing the network while maintaining the memory of the rule.

The Method of the Probability Analysis of Area with Dissolved Gases in Power Transformer Insulating Oil

Transformer insulating oil plays an extremely important role in power transformers and is used for insulation, arc suppression, cooling and other purposes. However, its other focus is on dissolved gases analysis in the oil to monitor internal operating conditions, which is the responsibility of preventive maintenance. This paper uses the normal distribution theory and the American National Standard Specification (hereinafter referred to as ANSI/IEEE C57.104) to reorganize as a diagnostic tool. The range of each gas from three stages of the specification - normal, caution, and abnormal are incorporated to a stage of abnormal, which divide into equal different values of 1000 as the maternal body of a qualitative normal distribution. Then the benchmark is been calculated from those parameters of a qualitative normal distribution. The data of detection has to pass through "Gas chromatography” to generate dissolved gases. At last, the benchmark value compares with dissolved gases to find the abnormal probability value as a diagnostic tool in maintenance evaluation of equipment. This method was developed using EXCEL application software, and repeatedly tested and verified with examples to confirm its feasibility and high accuracy. In addition, 10 cases accident data were extracted from the literature [11], and compared with the actual maintenance results. The accuracy was 80% for ANSI/IEEE C57.104 and 90% for this probability method. Therefore, this diagnostic method can replace the traditional multi-stage judgment and be represented by a single probability value to show. Based on technology sharing, the development process is specially written into a technical article as reference by scholars and electrical maintenance personnel in the field of power engineering.
 Design Methods: Making a qualitative normal distribution employs EXCEL software and the discriminative rule of ANSI/IEEE C57.104. Design Purpose: As a diagnostic tool with dissolved gases in insulating oil. Design Effectiveness: Probability Method is better than traditional; it has those efficacies - the accuracy and feasibly and compete with tried and tested actual case.

Recognition of Glaucomatous Fundus Images Using Machine Learning Methods Based on Optic Nerve Head Topographic Features.

Machine learning classifiers are an effective approach to detecting glaucomatous fundus images based on optic disc topographic features making it a straightforward and effective approach. Retrospective case-control study. The aim was to compare the effectiveness of clinical discriminant rules and machine learning classifiers in identifying glaucomatous fundus images based on optic disc topographic features. The study used a total of 800 fundus images, half of which were glaucomatous cases and the other half non-glaucomatous cases obtained from an open database and clinical work. The images were randomly divided into training and testing sets with equal numbers of glaucomatous and non-glaucomatous images. An ophthalmologist framed the edge of the optic cup and disc, and the program calculated five features, including the vertical cup-to-disc ratio and the width of the optic rim in four quadrants in pixels, used to create machine learning classifiers. The discriminative ability of these classifiers was compared with clinical discriminant rules. The machine learning classifiers outperformed clinical discriminant rules, with the extreme gradient boosting method showing the best performance in identifying glaucomatous fundus images. Decision tree analysis revealed that the cup-to-disc ratio was the most important feature for identifying glaucoma fundus images. At the same time, the temporal width of the optic rim was the least important feature. Machine learning classifiers are an effective approach to detecting glaucomatous fundus images based on optic disc topographic features and integration with an automated program for framing and calculating the required parameters would make it a straightforward and effective approach.

Risk assessment of crane operation hazards using modified FMEA approach with Z-number and set pair analysis

In view of two key deficiencies of failure mode and effects analysis (FMEA) in crane operation hazard risk assessment—information loss and hazard grade lack, a crane operation hazard risk assessment model based on the Z-number and set pair analysis (Z-SPA) is proposed. In this research, the Z-number is used to address the information uncertainty. To reduce the risk losing evaluation information in the conversion of the Z-number into a crisp number, it is translated into an interval number. In this assessment model, the grades of crane operation hazards are determined using grade discrimination rules and a score function of a connection number based on set pair analysis. Two case studies are presented to exemplify the effectiveness of the approach. The first case validates the proposed method, and the second case illustrates the application of the modified FMEA to crane operation hazard risk analysis. Furthermore, the assessment results are analyzed and compared with previous studies to highlight the superiority of the proposed model.

Assessment of functional diversities in patients with Asthma, COPD, Asthma-COPD overlap, and Cystic Fibrosis (CF).

The objectives of the present study were to evaluate the discriminating power of spirometric and plethysmographic lung function parameters to differenciate the diagnosis of asthma, ACO, COPD, and to define functional characteristics for more precise classification of obstructive lung diseases. From the databases of 4 centers, a total of 756 lung function tests (194 healthy subjects, 175 with asthma, 71 with ACO, 78 with COPD and 238 with CF) were collected, and gradients among combinations of target parameters from spirometry (forced expiratory volume one second: FEV1; FEV1/forced vital capacity: FEV1/FVC; forced expiratory flow between 25-75% FVC: FEF25-75), and plethysmography (effective, resistive airway resistance: sReff; aerodynamic work of breathing at rest: sWOB), separately for in- and expiration (sReffIN, sReffEX, sWOBin, sWOBex) as well as static lung volumes (total lung capacity: TLC; functional residual capacity: FRCpleth; residual volume: RV), the control of breathing (mouth occlusion pressure: P0.1; mean inspiratory flow: VT/TI; the inspiratory to total time ratio: TI/Ttot) and the inspiratory impedance (Zinpleth = P0.1/VT/TI) were explored. Linear discriminant analyses (LDA) were applied to identify discriminant functions and classification rules using recursive partitioning decision trees. LDA showed a high classification accuracy (sensitivity and specificity > 90%) for healthy subjects, COPD and CF. The accuracy dropped for asthma (~70%) and even more for ACO (~60%). The decision tree revealed that P0.1, sRtot, and VT/TI differentiate most between healthy and asthma (68.9%), COPD (82.1%), and CF (60.6%). Moreover, using sWOBex and Zinpleth ACO can be discriminated from asthma and COPD (60%). Thus, the functional complexity of obstructive lung diseases can be understood, if specific spirometric and plethysmographic parameters are used. Moreover, the newly described parameters of airway dynamics and the central control of breathing including Zinpleth may well serve as promising functional marker in the field of precision medicine.

The dynamic strategy shifting task: Optimisation of an operant task for assessing cognitive flexibility in rats.

Although schizophrenia is associated with a broad range of symptoms including hallucinations, delusions, and reduced motivation, measures of cognitive dysfunction, including cognitive flexibility and executive function, are the strongest predictors of functional outcomes. Antipsychotic medications are useful for reducing psychotic symptoms, but they are ineffective at improving cognitive deficits. Despite extensive investment by industry, the transition from preclinical to clinical trials has not been successful for developing precognitive medications for individuals with schizophrenia. Here, we describe the optimisation of a novel dynamic strategy shifting task (DSST) using standard operant chambers to investigate the optimal stimuli required to limit the extensive training times required in previous tasks. We determined that optimal learning by male and female Sprague Dawley rats for the flexibility task incorporated dynamic strategy shifts between spatial rules, such as following a visual cue or responding at one location, and non-spatial rules, such as responding to a central visual or auditory cue. A minimum of 6 correct consecutive responses were required to make a within-session change in the behavioural strategies. As a proof of concept, we trained and tested 84 Sprague Dawley rats on the DSST, and then assessed their cognitive flexibility using a within-subject design after an acute dose of ketamine (0, 3, 10 mg/kg). Rats made fewer premature and more perseverant responses to initiate a trial following ketamine. The effects of ketamine on trials to criterion was dependent on the rule. Ketamine induced a significant improvement on the reversal of a non-spatial visual discrimination rule. There was no significant effect of ketamine on the spatial visual or response discrimination rules. The DSST is a novel assay for studying distinct forms of cognitive flexibility and offers a rapid and adaptable means of assessing the ability to shift between increasingly challenging rule conditions. The DSST has potential utility in advancing our understanding of cognitive processes and the underlying neurobiological mechanisms related to flexibility in neuropsychiatric and neurological conditions where executive dysfunctions occur.>.

Reversing Aging: Decline in Complex Olfactory Learning Can be Rectified by Restoring Intrinsic Plasticity of Hippocampal CA1 Pyramidal Neurons.

The acquisition of complex rules requires modifications in intrinsic plasticity of excitatory neurons within relevant brain areas. Olfactory discrimination (OD) rule learning occludes slow calcium-dependent potassium current (sIAHP ) in piriform cortex (PC) pyramidal neurons, which increases their intrinsic neuronal excitability. Similar learning-induced sIAHP changes are demonstrated in hippocampal CA1. The shutdown of sIAHP is mediated by the metabotropic activation of the kainate subtype glutamatergic receptor, GluK2. Here, the duration of training required for OD rule learning increased significantly as the mice matured and aged is first shown, which appears earlier in 5xFAD mice. At the cellular biophysical level, aging is accompanied by reduction in the post-burst AHP in these neurons, while neuronal excitability remains stable. This is in contrast to aging CA1 neurons that exhibit enhanced post-burst AHPs in previous reports. Kainate reduces post-burst AHP in adults, but not in aged PC neurons, whereas it reduces post-burst AHPs in hippocampal CA1 pyramidal neurons of both young and aged mice. Overexpression of GluK2 in CA1 neurons restores OD learning capabilities in aged wild-type and 5xFAD mice, to a level comparable to young adults. Activation of GluK2 receptors in selectively vulnerable neurons can prevent aging-related cognitive decline is suggested.

Polygenic risk scores for cervical HPV infection, neoplasia and cancer show potential for personalised screening: comparison of two methods

The era of precision medicine requires the achievement of accurate risk assessment. Polygenic risk scores (PRSs) have strong potential for increasing the benefits of nationwide cancer screening programs. The current pool of evidence on the role of a PRS as a risk stratification model in actual practice and implementation is limited. To better understand the impact of possible method-induced variance, we constructed and validated two PRSs for cervical cancer (CC) using the Estonian Biobank female population (691 CC cases and 13,820 controls) and evaluated their utility in predicting incident cervical neoplasia (CIN), cancer, and human papillomavirus (HPV) infection using two methods (LDPred and BayesRR-RC). This study demonstrated that two genetic risk scores were significantly associated with CIN, CC, and HPV infection incidence. Independent of the method, we demonstrated that women with elevated PRS values reached the observed cumulative risk levels of CIN or CC much earlier. Our results indicated that the PRS-based discrimination rules could differ substantially when the PRSs contain similar predictive information. In summary, our analysis indicated that PRSs represent a personalized genetic component that could be an additional tool for cervical cancer risk stratification, and earlier detection of abnormalities provides invaluable information for those at high risk.

A Quantum-Inspired Direct Learning Strategy for Positive and Unlabeled Data

Learning from only positive and unlabeled (PU) data has broad applications in fields such as web data mining, product recommendations and medical diagnosis, which aims to train a binary classifier in the absence of negative labeled data. However, due to the lack of negative label information, prevailing PU learning methods usually rely on prior knowledge of unknown class distributions heavily. In fact, without additional constraints imposed by the prior knowledge, a direct learning strategy to coordinate the underlying clustering information in unlabeled data with the label information from positive training data is often considered challenging. To tackle this challenge, we propose a direct PU learning strategy using quantum formalization. By employing neural networks as backends, the samples are mapped into two-qubit composite systems, which should be understood here as mathematical entities encapsulating various classical distributions of two classical bits. Subsequently, the two qubits within the systems are trained to be as independent as possible from each other, capturing patterns of different classes. At the same time, their measurement results serving as the model outputs are encouraged to be maximally dissimilar. These characteristics enable effective training of classifiers on PU data. After formulating an appropriate discriminant rule, we introduce a quantum-inspired PU method named qPU using the direct learning strategy. This method not only has the potential to alleviate parameter sensitivity issues caused by prior estimation in other methods but is also straightforward to implement. Finally, experiments conducted on 13 classical datasets validate the effectiveness of qPU.

Improved multi-search strategy A* algorithm to solve three-dimensional pipe routing design

Pipe routing design (PRD) has a significant impact on the safety and operational efficiency of the system. PRD is involved in multi-disciplinary knowledge such as mechanical engineering, computer graphics and mathematical geometry, and it can be abstracted as a complex combined optimization problem. A* algorithm is one of the most commonly employed approaches to solve two-dimensional (2-D) path planning. However, due to the central idea of A* algorithm is all-traversal search, the search process considers all possibilities and requires a blind search one by one, it is difficult to solve PRD problems in complex three-dimensional (3-D) space environment. In this study, an improved multi-search strategy A* algorithm (MSSA*) is presented to overcome the disadvantages of the traditional A* algorithm in solving 3-D PRD problems. The MSSA* algorithm contains four novel mechanisms which are the node directional discrimination rule, the double-layer domain extension search strategy, the multi-factor heuristic evaluation function, and the dynamic adaptive weighting strategy. Then, a range of experiments are conducted to prove the validity of the presented mechanisms. Subsequently, the MSSA* is compared with several existing algorithms for solving 3-D PRD to further verify the superiority of MSSA* in terms of high-efficiency and practicality. Finally, the MSSA* is applied to solve the 3-D PRD problems for offshore production platforms in oil and gas treatment system.

Concise and interpretable multi-label rule sets

Multi-label classification is becoming increasingly ubiquitous, but not much attention has been paid to interpretability. In this paper, we develop a multi-label classifier that can be represented as a concise set of simple “if-then” rules, and thus, it offers better interpretability compared to black-box models. Notably, our method is able to find a small set of relevant patterns that lead to accurate multi-label classification, while existing rule-based classifiers are myopic and wasteful in searching rules, requiring a large number of rules to achieve high accuracy. In particular, we formulate the problem of choosing multi-label rules to maximize a target function, which considers not only discrimination ability with respect to labels, but also diversity. Accounting for diversity helps to avoid redundancy, and thus, to control the number of rules in the solution set. To tackle the said maximization problem, we propose a 2-approximation algorithm, which circumvents the exponential-size search space of rules using a novel technique to sample highly discriminative and diverse rules. In addition to our theoretical analysis, we provide a thorough experimental evaluation and a case study, which indicate that our approach offers a trade-off between predictive performance and interpretability that is unmatched in previous work.

Incipient fault detection and process monitoring of thermal power plant pulverizing system based on deep representation learning

Process data in contemporary thermal power plants shows the characteristics of large capacity, strong coupling, and high-order nonlinearity, which brings great challenges to process monitoring and fault detection. A deep representation learning fault detection scheme based on stacked sparse denoising autoencoder (SSDAE) is proposed in this paper. Specifically, to enhance the capabilities of noise reduction and feature representation for complex industrial data, the sparse denoising autoencoder (SDAE) is proposed by considering both noise and sparsity constraints. Then, a deep learning architecture is constructed by stacking multiple SDAEs layer by layer to achieve a highly nonlinear representation capability. Based on the low-dimensional representation and residual distance of SSDAE, three monitoring indicators, RE2, MD2, and ZD2, are designed by different distance metrics and the k-nearest neighbor (KNN) discriminant rule. The effectiveness of the proposed method is validated by studying a nonlinear numerical case and a practical power plant pulverizing system. The experimental results demonstrate that the proposed method can effectively detect incipient and slight faults that are difficult to detect with traditional methods.

Classification of objects within agriculture landscape from high resolution aerial imagery using maximum likelihood discriminant rules

In this study, the performance of object classification based on four discriminant functions, namely linear, quadratic, diagonal linear and diagonal quadratic is investigated and compared. High-resolution aerial imagery captured from a UAV-based remote sensing platform is used for this purpose. Initially, K-means clustering of 9 clusters is used to assist in the selection of training pixels for the subsequent supervised classification implementation. The classification is experimented with using a training set size of 10 through 100 pixels for each of the discriminant functions. The outcome of the classification shows that training set size 40 and 10 are to be the optimal training set sizes for linear and quadratic discriminant function, and diagonal linear and quadratic discriminant function respectively. Overall, the linear discriminant function is found to have the highest overall accuracy of 91% followed by diagonal linear, quadratic, and diagonal quadratic discriminant function with overall accuracies of 82%, 79%, and 73% respectively.

Incrementally Mining Column Constant Biclusters with FVSFP Tree

Bicluster mining has been frequently studied in the data mining field. Because column constant biclusters (CCB) can be transformed to be discriminative rules, they have been widely applied in various fields. However, no research on incrementally mining CCB has been reported in the literature. In real situations, due to the limitation of computation resources (such as memory), it is impossible to mine biclusters from very large datasets. Therefore, in this study, we propose an incremental mining CCB method. CCB can be deemed as a special case of frequent pattern (FP). Currently the most frequently used method for incrementally mining frequent patterns is FP tree based method. In this study, we innovatively propose an incremental mining CCB method with modified FP tree data structure. The technical contributions lie in two aspects. The first aspect is that we propose a modified FP tree data structure, namely Feature Value Sorting Frequent Pattern (FVSFP) tree that can be easily maintained. The second aspect is that we innovatively design a method for mining CCB from FVSFP tree. To verify the performance of the proposed method, it is tested on several datasets. Experimental results demonstrated that the proposed method has good performance for incrementally handling a newly added dataset.

DAC: Discriminative Associative Classification

In this paper, discriminative associative classification is proposed as a new classification technique based on class discriminative association rules (CDARs). These rules are defined based on discriminative itemsets. The discriminative itemset is frequent in one data class and has much higher frequencies compared with the same itemset in other data classes. The CDAR is a class associative rule (CAR) in one data class that has higher support compared with the same rule in other data classes. Compared to associative classification, there are additional challenges as the Apriori property of the subset is not applicable. The proposed algorithm is designed particularly based on well-defined distinguishing characteristics of the rules, to improve the accuracy and efficiency of the classification in data classes. A novel compact prefix-tree structure is defined for holding the rules in data classes. The empirical analysis shows the effectiveness and efficiency of the proposed method on small and large real datasets.

Who earns more? City hierarchy of earning returns to education among migrant workers in urban China

According to urban economic theories, workers in more developed regions should have higher educational returns. However, studies on urban workers in China have shown mixed results on whether earning returns to education are higher in regions experiencing more economic growth. Meanwhile, there is a lack of evidence on this issue among migrant populations, which have rapidly increased in number during China's economic transition. To fill this void, this study investigated how educational returns among migrant workers are related to local economic development levels and industrial structures. Specifically, we used individual-level data obtained from 286 cities through the 2018 China Migrants Dynamic Survey and city-level data to conduct a multilevel linear regression analysis with random slopes. We thus found that earning returns to education were higher in cities under two conditions, including higher development and higher shares of tertiary industry. Moreover, this association was more pronounced among high-skill migrant workers than low-skill ones. As higher income is a primary driving force in migration, we suggest that future policies should further promote industrial restructuring rather than using discriminative administrative rules to guide migration flows.

Research on glass classification based on linear regression model

This paper studies the rule of chemical composition of ancient glass products and the identification of glass types. Through Spearman correlation analysis, the establishment of multiple linear regression model and Fisher linear discriminant model, the relationship and classification rule of chemical composition of different types of glass are determined, and the linear regression model of chemical composition content before and after weathering and the identification method of glass types are obtained. The results show that the mathematical model proposed in this paper can effectively identify glass types and provide guidance for the identification of glass components.

A Process-Oriented Exploration of the Evolutionary Structures of Ocean Dynamics with Time Series of a Remote Sensing Dataset

Advanced Earth observation technologies provide a tool for the study of ocean dynamics either in basins or in oceans. In a comparison of when and where, how ocean dynamics evolves in space and time is still a challenge. In view of an evolutionary scale, this paper proposes a novel approach to explore the evolutionary structures of ocean dynamics with time series of a raster dataset. This method, called PoEXES, includes three key steps. Firstly, a cluster-based algorithm is enhanced by process semantics to obtain marine snapshot objects. Secondly, the discriminant rule is formulated on the basis of successive marine snapshot objects’ spatiotemporal topologies to identify marine sequence objects and marine linked objects. Thirdly, a sequence-linked object-based algorithm (SLOA) is used for marine sequence objects and linked objects to obtain their evolutionary structures and to achieve four evolutionary relationships, i.e., development, merging, splitting, and a splitting–merging relationship. Using the evolutionary relationships and their occurring orders in a lifespan of ocean dynamics, this paper reformulates five types of evolutionary structures, which consist of origination nodes, linked nodes, sequence nodes and dissipation nodes. The evolutionary-scale-based dynamic structure ensures the optimum evolutionary relationships of ocean dynamics as much as possible, which provides a new way to design a spatiotemporal analysis model for dealing with geographical dynamics. To demonstrate the effectiveness and the advantages of PoEXES, a real 40-year dataset of satellite-derived sea surface temperatures is used to explore the evolutionary structure in global oceans; the new findings may help to better understand global climate change.

Liability Rules for Algorithmic Discrimination

Liability Rules for Algorithmic Discrimination