Boost-wise pre-loaded mixture of experts for classification tasks

A modified version of Boosted Mixture of Experts (BME) is presented in this paper. While previous related works, namely BME, attempt to improve the performance by incorporating complementary features of a hybrid combining framework, they have some drawback. Analyzing the problems of previous approaches has suggested several modifications that have led us to propose a new method called Boost-wise Pre-loaded Mixture of Experts (BPME). We present a modification in pre-loading (initialization) procedure of ME, which addresses previous problems and overcomes them by employing a two-stage pre-loading procedure. In this approach, both the error and confidence measures are used as the difficulty criteria in boost-wise partitioning of problem space.

     

A Reservoir Conductivity Evaluation Using Percolation Theory

Abstract

Oil reservoirs are very complex with geological heterogeneities that appear on all scales. Proper modeling of the spatial distribution of these heterogeneities is crucial, affecting all aspects of flow and, consequently, the reservoir performance. Reservoir connectivity and conductivity evaluation is of great importance for decision-making on various possible development scenarios including infill drilling projects. This can be addressed by using the percolation theory approach. This statistical approach considers a hypothesis that the reservoir can be split into either permeable (good sands) or impermeable flow units (poor sands) and assumes that the continuity of permeability contrasts controls the flow. The approach uses an object-based technique to model the spatial distribution of both isotropic and anisotropic sand bodies in two dimensions. Recently the application of the percolation approach in evaluating reservoir connectivity for both conventional and fracture reservoirs has been reported. This article concentrates on analyzing the conductivity behavior of petroleum reservoirs. In particular, the percolation approach is used to develop the universal curves of the average reservoir permeability and its associated uncertainties.

To validate the approach, we used the Burgan reservoir dataset of Norouz offshore oil field in the south of Iran. Core and palynological data analysis indicated that the Burgan consists of a series of incision-fill sequences occurring in an estuarine/coastal plain/deltaic environment. Consequently, the Burgan consists of a thick stack of excellent quality sands incising into each other with few remaining shalier sediments locally separating these sequences. We have shown that the effective permeability, as a measure of reservoir conductivity, from the percolation approach gives reliable prediction once compared with exact numerically obtained results from the modeling of real field data. Moreover, the first approach as obtained from algebraic manipulation is very fast whereas the latter is so costly and time consuming.     

Electrocardiogram beat classification via coupled boosting by filtering and preloaded mixture of experts

The role of electrocardiogram (ECG) as a noninvasive technique for detecting and diagnosing cardiac problems cannot be overemphasized. This paper introduces a modified version of boosted mixture of experts for the classification of three types of ECG beats. Our two-step preloading procedure, along noise injection, also regarded as smoothing regularization, proved to be a promising, effective, and safe means of classifying arrhythmias. The proposed model, according to the nature of implementation, is called coupled boosting by filtering and preloaded mixture of experts. The experimental results show our proposed method have better classification rate against other compared methods. Comparative evaluation is accomplished with ECG signals from MIT–BIH arrhythmia database.