Relevance Vector Machine Approach Research Articles

Harvesting Brain Signal Using Machine Learning Methods

Abstract Brain computer interface (BCI) systems are developed in the biomedical engineering fields to increase the quality of life among patients with paralysis and neurological conditions. The development of a six class BCI controller to operate a semi-autonomous mobile robotic arm is presented. The controller uses the following mental tasks: imagined left/right hand squeeze, imagined left/right foot tap, rest, and a physical jaw clench. To design a controller, the locations of active electrodes are verified, and an appropriate machine learning algorithm is determined. Three subjects, ages ranging between 22 and 27, participated in five sessions of motor imagery experiments to record their brainwaves. These recordings were analyzed using event related potential (ERP) plots and topographical maps to determine active electrodes. bcilab was used to train two, three, five, and six class BCI controllers using linear discriminant analysis (LDA) and relevance vector machine (RVM) machine learning methods. The subjects' data were used to compare the two-method's performance in terms of error rate percentage. While a two class BCI controller showed the same accuracy for both methods, the three and five class BCI controllers showed the RVM approach having a higher accuracy than the LDA approach. For the five-class controller, error rate percentage was 33.3% for LDA and 29.2% for RVM. The six class BCI controller error rate percentage for both LDA and RVM was 34.5%. While the percentage values are the same, RVM was chosen as the desired machine learning algorithm based on the trend seen in the three and five class controller performances.

Executive functions predict verbal fluency scores in healthy participants

While there is a clear link between impairments of executive functions (EFs), i.e. cognitive control mechanisms that facilitate goal-directed behavior, and speech problems, it is so far unclear exactly which of the complex subdomains of EFs most strongly contribute to speech performance, as measured by verbal fluency (VF) tasks. Furthermore, the impact of intra-individual variability is largely unknown. This study on healthy participants (n = 235) shows that the use of a relevance vector machine approach allows for the prediction of VF performance from EF scores. Based on a comprehensive set of EF scores, results identified cognitive flexibility and inhibition as well as processing speed as strongest predictors for VF performance, but also highlighted a modulatory influence of fluctuating hormone levels. These findings demonstrate that speech production performance is strongly linked to specific EF subdomains, but they also suggest that inter-individual differences should be taken into account.

(Invited) Using Machine Learning to Decode Output of a Mixed-Potential Sensor Array for Automotive Exhaust Monitoring

Introduction In order to meet the latest requirements for more fuel-efficient cars with low emissions, it is essential to equip vehicles with advanced sensor systems whose readouts can be used to determine the content of key automotive exhaust pollutants such as NOx, NH3 and C3H8 in real time [1]. Arrays of mixed-potential electrochemical sensors (MPES) provide a competitive sensor platform for this task. MPES are electrochemical devices that develop a non-Nernstian potential due to differences in the redox kinetics of various gas species at each electrode/electrolyte gas interface [2,3]. To mitigate signal drift and poor reproducibility, we have developed a patented MPES design utilizing dense electrodes with an overcoat of porous electrolytes such as yttria-stabilized zirconia (YSZ). This design has resulted in highly stable, reproducible, and durable devices which were tested for up to 1000h of operation [4]. In this study, we have employed an array of four mixed-potential sensors (Cr450, Au475, H545, and Cr470) to monitor gas mixtures that mimic automotive exhaust in a laboratory setting.However, even with the latest sensor designs we observe significant cross-specificity and non-linearity in sensor responses to each target gas. Thus, reliable inference of the concentration (or even the presence) of each constituent gas in a complex mixture requires application of advanced machine-learning techniques that will be the focus of this presentation. The ability to decipher the content of gas or liquid mixtures both quickly and reliably has potential applications in many areas of science and technology, including monitoring of various technological processes and continuous observation of air quality in the interests of ecological studies and national security. Our computational methodology can be readily adapted to various sensor array platforms in which gas or liquid mixtures elicit complex response patterns. Method Our previous work has focused on developing computational models that were based on the fundamental electrochemistry of MPES and utilized detailed quantitative descriptions of gas-sensor interactions [5]. We were able to treat both two- and three-gas mixtures (Fig. 1) and take into account the non-linearities of sensor responses. Our model was able to estimate relative concentrations of C3H8, NH3 and NO2 with respect to NO with the maximum error of 14.0% and the average error of 1.8%. Furthermore, we predicted the absolute concentration of each gas in two- and three-gas mixtures with the average error of 3.2%. In this presentation, we discuss application of a general-purpose machine-learning technique called Bayesian regression to sensor array data. The primary advantages of this technique are its versatility (that is, the technique is relatively independent of the sensor platform employed to provide the measurements) and, given a novel measurement, its ability to estimate both the expected concentration of each component in the mixture and the corresponding uncertainty of the predictions. Unlike maximum-likelihood methods, Bayesian regression does not suffer from overfitting and readily lends itself to model comparison, allowing us to choose the optimal model complexity given the available data. Finally, Bayesian regression can be treated using the relevance vector machine approach, which is specifically designed to produce sparse, interpretable solutions. Discussion and Conclusions There is a pressing need for a computational approach designed to infer gas concentrations in complex mixtures such as diesel exhaust on the basis of sensor array outputs. In this presentation, we will discuss Bayesian regression analysis of two-, three- and four-gas mixtures (Fig. 1) using voltage readouts of an array of four mixed-potential sensors as input. Our Bayesian approach is scalable, robust, and easily transferrable to other sensor platforms that are currently employed in numerous scientific and industrial settings. After fitting the model parameters on gas-mixture data generated in the lab under controlled conditions, the resulting model can be employed to predict gas concentrations from novel measurements in real time, as required by real-world applications. Figure 1. Plot of the voltage response (in volts) of three sensors in the array: Cr450, Au475 and H545 (Cr470 was excluded due to its similarity with Cr450). Color coding indicates mixture type.

Daily sea level prediction at Chiayi coast, Taiwan using extreme learning machine and relevance vector machine

Abstract The analysis and the prediction of sea level fluctuations are core requirements of marine meteorology and operational oceanography. Estimates of sea level with hours-to-days warning times are especially important for low-lying regions and coastal zone management. The primary purpose of this study is to examine the applicability and capability of extreme learning machine (ELM) and relevance vector machine (RVM) models for predicting sea level variations and compare their performances with powerful machine learning methods, namely, support vector machine (SVM) and radial basis function (RBF) models. The input dataset from the period of January 2004 to May 2011 used in the study was obtained from the Dongshi tide gauge station in Chiayi, Taiwan. Results showed that the ELM and RVM models outperformed the other methods. The performance of the RVM approach was superior in predicting the daily sea level time series given the minimum root mean square error of 34.73 mm and the maximum determination coefficient of 0.93 (R2) during the testing periods. Furthermore, the obtained results were in close agreement with the original tide-gauge data, which indicates that RVM approach is a promising alternative method for time series prediction and could be successfully used for daily sea level forecasts.

A Modified Relevance Vector Machine for PEM Fuel-Cell Stack Aging Prediction

Proton exchange membrane fuel cells (PEMFCs) are considered as a potential candidate in the green-energy applications in the near future. Comparing with other energy options, the PEMFCs need only hydrogen and air during operation. Meanwhile, as a by-product during operation, water is produced. This energy-conversion process is 100% eco-friendly and completely unharmful to the environment. However, PEMFCs are vulnerable to the impurities of hydrogen or fluctuation of operational condition, which could cause the degradation of output performance over time during operation. Thus, the prediction of the performance degradation is critical to the PEMFC system. In this work, a novel PEMFC performance-forecasting model based on a modified relevance vector machine (RVM) has been proposed, followed by a comparison with the approach of classic support vector machine (SVM). First, the theoretical formulation of RVM is briefly introduced, then the implementation steps of RVM using the experimental aging data sets of PEMFC stack output voltage are presented. By considering the specific feature of aging data-prediction problem, an innovative modified RVM formulation is proposed. The results of proposed modified RVM method are analyzed and compared to the results of SVM. The results have demonstrated that the modified RVM can achieve better performance of prediction than SVM, especially in the cases with relatively small training data sets. This novel method based on modified RVM approach has been demonstrated to show its effectiveness on forecasting the performance degradation of PEMFCs.

Reconstruction of multiple impact forces by wavelet relevance vector machine approach

In this work, reconstruction of forces generated by multiple impacts occurring on linear elastic structures has been achieved through wavelet relevance vector machine approach to inverse problem solution. The posterior density of probabilities function integrating both the likelihood and prior random information was considered for the particular case of a system being affected by an additive random measurement noise. Compressed sensing technique was used to provide an adequate sparse representation of the inverse problem through using projection on a wavelet basis. Bayesian hierarchical modeling according to the relevance vector machine approximation was then applied in order to estimate the forces generated by impact events. The obtained results were remarkably good as the reconstructed forces were found to be very close to the original forces at the system input. The method was found also to provide a way for localization of impact forces.

Predicting hydrofacies and hydraulic conductivity from direct-push data using a data-driven relevance vector machine approach: Motivations, algorithms, and application

The spatial heterogeneity of hydraulic conductivity (K) exerts a major control on groundwater flow and solute transport. The heterogeneous spatial distribution of K can be imaged using indirect geophysical data as long as reliable relations exist to link geophysical data to K. This paper presents a nonparametric learning machine approach to predict aquifer K from cone penetrometer tests (CPT) coupled with a soil moisture and resistivity probe (SMR) using relevance vector machines (RVMs). The learning machine approach is demonstrated with an application to a heterogeneous unconsolidated littoral aquifer in a 12 km2 subwatershed, where relations between K and multiparameters CPT/SMR soundings appear complex. Our approach involved fuzzy clustering to define hydrofacies (HF) on the basis of CPT/SMR and K data prior to the training of RVMs for HFs recognition and K prediction on the basis of CPT/SMR data alone. The learning machine was built from a colocated training data set representative of the study area that includes K data from slug tests and CPT/SMR data up-scaled at a common vertical resolution of 15 cm with K data. After training, the predictive capabilities of the learning machine were assessed through cross validation with data withheld from the training data set and with K data from flowmeter tests not used during the training process. Results show that HF and K predictions from the learning machine are consistent with hydraulic tests. The combined use of CPT/SMR data and RVM-based learning machine proved to be powerful and efficient for the characterization of high-resolution K heterogeneity for unconsolidated aquifers.

Riding a probabilistic support vector machine to the Stanley Cup

Abstract:The predictive performance of various team metrics is compared in the context of 105 best-of-seven national hockey league (NHL) playoff series that took place between 2008 and 2014 inclusively. This analysis provides renewed support for traditional box score statistics such as goal differential, especially in the form of Pythagorean expectations. A parsimonious relevance vector machine (RVM) learning approach is compared with the more common support vector machine (SVM) algorithm. Despite the potential of the RVM approach, the SVM algorithm proved to be superior in the context of hockey playoffs. The probabilistic SVM results are used to derive playoff performance expectations for NHL teams and identify playoff under-achievers and over-achievers. The results suggest that the Arizona Coyotes and the Carolina Hurricanes can both be considered Round 2 over-achievers while the Nashville Predators would be Round 2 under-achievers, even after accounting for several observable team performance metrics and playoff predictors. The Vancouver Canucks came the closest to qualify as Stanley Cup Finals under-achievers after they lost against the Boston Bruins in 2011. Overall, the results tend to support the idea that the NHL fields extremely competitive playoff teams, that chance or other intangible factors play a significant role in NHL playoff outcomes and that playoff upsets will continue to occur regularly.

Development of an adaptive relevance vector machine approach for slope stability inference

Uncertainty is commonly encountered in such problems as the stability inference of slopes in earth science and geotechnical engineering. This uncertainty can be approached by the artificial intelligence techniques and experts systems. This paper presents the adaptive relevance vector machine (ARVM) for stability inference of soil slopes. Based on failure mechanisms and due to data availability, the stability inference here is realized according to the three categories of slope parameters: (1) geomaterial parameters, (2) slope geometry parameters, and (3) the pore pressure coefficient R u . A database of dozens of slope cases is collected to reasonably execute the stability inference. Then, the ARVM is introduced to approach the problem. Some cases in the database are used to train the ARVM model so that an optimized ARVM model can be obtained. Some other cases are then used to test the inference ability of the optimized model. Four models obtained by different numbers of cases are compared to show possible effects of dataset size. Also, the sensitivity of the ARVM parameters is investigated. The results show that the width hyper-parameter has apparent effects on the performance of the ARVMs, and the kernel type as well as the dataset size can result in different optimal hyper-parameter values. Meanwhile, the ARVM is compared to other techniques such as the generalized regression neural network and the support vector machines (SVM) for the stability inference of the slope cases by the inference accuracy function. The results suggest that the ARVMs have satisfactory generalization ability and perform better than the simple SVM and the applied neural networks.

Prediction of elastic compressibility of rock material with soft computing techniques

Mechanical and physical properties of sandstone are interesting scientifically and have great practical significance as well as their relations to the mineralogy and pore features. These relations are however highly nonlinear and cannot be easily formulated by conventional methods. This paper investigates the potential of the technique named as the relevance vector machine (RVM) for prediction of the elastic compressibility of sandstone based on its characteristics of physical properties. Based on the fact that the hyper-parameters may have effects on the RVM performance, an iteration method is proposed in this paper to search for optimal hyper-parameter value so that it can produce best predictions. Also, the qualitative sensitivity of the physical properties is investigated by the backward regression analysis. Meanwhile, the hyper-parameter effect of the RVM approach is discussed in the prediction of the elastic compressibility of sandstone. The predicted results of the RVM demonstrate that hyper-parameter values have evident effects on the RVM performance. Comparisons on the results of the RVM, the artificial neural network and the support vector machine prove that the proposed strategy is feasible and reliable for prediction of the elastic compressibility of sandstone based on its physical properties.

SPT-based liquefaction potential assessment by relevance vector machine approach

This article adopts Relevance Vector Machine (RVM) for the determination of the liquefaction potential of soil based on Standard Penetration Test (SPT) data from Chi-Chi earthquake. RVM is a probabilistic sparse kernel model. This study uses RVM for solving binary classification problem. Two models (MODEL I and MODEL II) have been developed. Cyclic Stress Ratio and SPT blow count (N) have been used as input variables for MODEL I. MODEL II uses Peak Ground Acceleration (PGA) and N as input variables. The developed RVM model gives equations for the prediction of the liquefaction potential of soil. A comparative study has been presented between the developed RVM and the Artificial Neural Network (ANN). The results confirm that the developed RVM is a robust model for the prediction of the liquefaction potential of soil.

ECG Arrhythmia Detection and Classification Using Relevance Vector Machine

The Electrocardiogram (ECG) is one of the most effective diagnostic tools to detect cardiac diseases. It is a method to measure and record different electrical potentials of the heart. The electrical potential generated by electrical activity in cardiac tissue is measured on the surface of the human body. This ECG can be classified as normal and abnormal signals. In this paper, a thorough experimental study was conducted to show the superiority of the generalization capability of the Relevance Vector Machine (RVM) in the automatic classification of ECG beats. To achieve the maximum accuracy the RVM classifier design by searching for the best value of the parameters that tune its discriminant function, and upstream by looking for the best subset of features that feed the classifier. The experiments were conducted on the ECG data from the Massachusetts Institute of Technology–Beth Israel Hospital (MIT–BIH) arrhythmia database to classify five kinds of abnormal waveforms and normal beats. The obtained results clearly confirm the superiority of the RVM approach when compared to traditional classifiers.

Pullout capacity of small ground anchors: a relevance vector machine approach

This paper examines the potential of relevance vector machine (RVM) in prediction of pullout capacity of small ground anchors. RVM is based on a Bayesian formulation of a linear model with an appropriate prior that results in a sparse representation. The results are compared with a widely used artificial neural network (ANN) model. Overall, the RVM showed good performance and is proven to be better than ANN model. It also estimates the prediction variance. The plausibility of RVM technique is shown by its superior performance in forecasting pullout capacity of small ground anchors providing exogenous knowledge.

APPLICATION OF RELEVANCE VECTOR MACHINE IN SEISMIC ATTENUATION PREDICTION

The recently introduced relevance vector machine (RVM) technique is applied to predict seismic attenuation based on rock properties. The RVM provides much sparser regressors without compromising performance, and kernel bases give a small but worthwhile improvement in performance. It evades complexity by producing models that have structure and as a result parameterization process that is appropriate to the information content of the data. Sensitivity analysis has been also performed to investigate the importance of each of the input parameters. The results show that RVM approach has the potential to be a practical tool for determination of seismic attenuation.

Relevance vector machine for automatic detection of clustered microcalcifications

Clustered microcalcifications (MC) in mammograms can be an important early sign of breast cancer in women. Their accurate detection is important in computer-aided detection (CADe). In this paper, we propose the use of a recently developed machine-learning technique--relevance vector machine (RVM)--for detection of MCs in digital mammograms. RVM is based on Bayesian estimation theory, of which a distinctive feature is that it can yield a sparse decision function that is defined by only a very small number of so-called relevance vectors. By exploiting this sparse property of the RVM, we develop computerized detection algorithms that are not only accurate but also computationally efficient for MC detection in mammograms. We formulate MC detection as a supervised-learning problem, and apply RVM as a classifier to determine at each location in the mammogram if an MC object is present or not. To increase the computation speed further, we develop a two-stage classification network, in which a computationally much simpler linear RVM classifier is applied first to quickly eliminate the overwhelming majority, non-MC pixels in a mammogram from any further consideration. The proposed method is evaluated using a database of 141 clinical mammograms (all containing MCs), and compared with a well-tested support vector machine (SVM) classifier. The detection performance is evaluated using free-response receiver operating characteristic (FROC) curves. It is demonstrated in our experiments that the RVM classifier could greatly reduce the computational complexity of the SVM while maintaining its best detection accuracy. In particular, the two-stage RVM approach could reduce the detection time from 250 s for SVM to 7.26 s for a mammogram (nearly 35-fold reduction). Thus, the proposed RVM classifier is more advantageous for real-time processing of MC clusters in mammograms.