Sequential Selection Algorithm Research Articles

A new sequential homogeneous pixel selection algorithm for distributed scatterer InSAR

ABSTRACT Distributed scatterer interferometric synthetic aperture radar (DS InSAR) technology has been widely used in various fields. Homogeneous pixel selection is a crucial step in the use of DS InSAR, directly affecting the estimation precision and reliability of subsequent parameter calculations. The existing algorithms for selecting homogeneous pixels have inherent limitations, such as requiring many heterogeneous samples and strict requirements surrounding the required number of synthetic aperture radar (SAR) images. To address these problems, a new sequential selection algorithm for homogeneous pixels is proposed, based on the Baumgartner – Weiss–Schindler (BWS) test algorithm and dynamic interval estimation (DIE) theory. According to Monte Carlo simulation experiments, the average standard deviation (STD) of the mean of the rejection of the BWS-DIE algorithm under six sample conditions is 0.014. Compared with three existing algorithms, including the Kolmogorov‒Smirnov (KS), BWS and fast statistically homogeneous pixel selection (FaSHPS) algorithms, the BWS-DIE algorithm improves homogeneous pixel selection precision by 64.3%, 69.4% and 25.3%, respectively. In the real data experiment, 12 scenes of Advanced Land Observing Satellite-1 Phased Array type L-band Synthetic Aperture Radar (ALOS-1 PALSAR) data from February 2007 to March 2011 were used and the BWS-DIE multitemporal InSAR (MT InSAR) method based on the BWS-DIE algorithm was applied to surface subsidence monitoring in the western mining area of Xuzhou, Jiangsu Province, China. The experimental results show that, compared with the Stanford Method for Persistent Scatterers (StaMPS), the BWS-DIE MT InSAR method improves the ability to monitor the maximum subsidence by 12.3%, increases the point density by 5.7 times and decreases the root mean square error (RMSE) by 50%. In addition, new surface deformation patterns are found in the spatial-temporal evolution. The above experimental results show that the proposed BWS-DIE algorithm exhibits remarkable advantages in selection power and selection precision and is not limited by the number of SAR images. The proposed algorithm can further broaden the application scenarios for DS InSAR and provide high-quality and reliable monitoring data for subsequent scientific research.

Ensemble method based architecture using random forest importance to predict employee’s turn over

The departure of a skilled employee can create a problem for a company and this incident is increasing globally. Employee turnover has become an important issue these days due to the heavy workload, low pay, low job satisfaction, poor working environment. Companies face problems as their budget will increase, losing skilled manpower and employees’ trust. It’s taking time to adjust for a new employee and bring risk and increase the cost for the company. It is necessary to bring appropriate solutions to the problem. The main purpose of this paper is to predict the turnover of employees with the help of state of the art machine learning classifier. We have determined employee turnover selection factors using some prediction models. We first pre-processed the dataset by removing correlative attributes. Then, we have scaled the attributes. Secondly, a Sequential selection algorithm (SBS) has been using to reduce features from a high number to a relatively small signal-canton. Then use Chi-square and Random Forest important algorithms to determine the most significant shared key features. Then we get average_montly_hours, satisfaction_level, time_spend_company are responsible for the employee’s departure. Then, we have applied different state of the art machine learning algorithm to measure the accuracy. We have achieved the highest accuracy of 99.4% using the reduced feature with 10-Fold Cross-validation by applied the Random Forest classifier and which is higher than the mentioned reference work.

Density-based Sensor Selection for RSS Target Localization

In received signal strength target localization problems, the selection of sensor positions involved in estimation is a crucial factor in determining the network’s overall quality of estimation. In this paper, we consider a region of interest with: 1) several points of interest (PoIs) each assigned a localization error requirement and 2) candidate sensor positions at which sensors can be deployed/activated. Then, under sensor budget constraints, how can we select a limited number of sensors so that the overall localization error at the PoIs is best met in a squared error sense? Instead of directly using errors, we use corresponding sensor density information which results in the simplification of the selection problem. Using density information, we propose two novel sequential selection algorithms. In the first algorithm, selection is formulated as a sequential convex quadratic optimization problem in which the cost is a function of the difference between required and achieved sensor densities at the PoIs. Selection in the second algorithm uses density weights assigned to each candidate position and is performed in a greedy fashion. Performance evaluation of the proposed methods with respect to other algorithms shows their effectiveness.

Image feature selection using genetic programming for figure-ground segmentation

Figure-ground segmentation is the process of separating regions of interest from unimportant background. One challenge is to segment images with high variations (e.g. containing a cluttered background), which requires effective feature sets to capture the distinguishing information between objects and backgrounds. Feature selection is necessary to remove noisy/redundant features from those extracted by image descriptors. As a powerful search algorithm, genetic programming (GP) is employed for the first time to build feature selection methods that aims to improve the segmentation performance of standard classification techniques. Both single-objective and multi-objective GP techniques are investigated, based on which three novel feature selection methods are proposed. Specifically, one method is single-objective, called PGP-FS (parsimony GP feature selection); while the other two are multi-objective, named nondominated sorting GP feature selection (NSGP-FS) and strength Pareto GP feature selection (SPGP-FS). The feature subsets produced by the three proposed methods, two standard sequential selection algorithms, and the original feature set are tested via standard classification algorithms on two datasets with high variations (the Weizmann and Pascal datasets). The results show that the two multi-objective methods (NSGP-FS and SPGP-FS) can produce feature subsets that lead to solutions achieving better segmentation performance with lower numbers of features than the sequential algorithms and the original feature set based on standard classifiers for given segmentation tasks. In contrast, PGP-FS produces results that are not consistent for different classifiers. This indicates that the proposed multi-objective methods can help standard classifiers improve the segmentation performance while reducing the processing time. Moreover, compared with SPGP-FS, NSGP-FS is equally capable of producing effective feature subsets, yet is better at keeping diverse solutions.

Multiple robots formation manoeuvring and collision avoidance strategy

This paper presents a multiple robots formation manoeuvring and its collision avoidance strategy. The direction priority sequential selection algorithm is employed to achieve the raw path, and a new algorithm is then proposed to calculate the turningcompliant waypoints supporting the multi-robot formation manoeuvre. The collision avoidance strategy based on the formation control is presented to translate the collision avoidance problem into the stability problem of the formation. The extension-decompositionaggregation scheme is next applied to solve the formation control problem and subsequently achieve the collision avoidance during the formation manoeuvre. Simulation study finally shows that the collision avoidance problem can be conveniently solved if the stability of the constructed formation including unidentified objects can be satisfied.

Light Reflectance Spectroscopy to Detect Positive Surgical Margins on Prostate Cancer Specimens

Intraoperative frozen section analysis is not routinely performed todetermine positive surgical margins at radical prostatectomy due to time requirementsand unproven clinical usefulness. Light reflectance spectroscopy, which measures light intensity reflected or backscattered from tissues, can be applied to differentiate malignant from benign tissue. We used a novel light reflectance spectroscopy probe to evaluate positive surgical margins on exvivo radical prostatectomy specimens and correlate its findings with pathological examination. Patients with intermediate to high risk disease undergoing radical prostatectomy were enrolled. Light reflectance spectroscopy was performed on suspected malignant and benign prostate capsule immediately following organ extraction. Each light reflectance spectroscopy at 530 to 830 nm was analyzed and correlated with pathological results. A regression model and forward sequential selection algorithm were developed for optimal feature selection. Eighty percent of light reflectance spectroscopy data were selected to train a logistic regression model, which was evaluated by the remaining 20% data. This was repeated 5 times to calculate averaged sensitivity, specificity and accuracy. Light reflectance spectroscopy analysis was performed on 17 exvivo prostate specimens, on which a total of 11 histologically positive and 22 negative surgical margins were measured. Two select features from 700 to 830 nm were identified as unique to malignant tissue. Cross-validation when performing thepredictive model showed that the optical probe predicted positive surgical margins with 85% sensitivity, 86% specificity, 86% accuracy and an AUC of 0.95. Light reflectance spectroscopy can identify positive surgical margins accurately in fresh exvivo radical prostatectomy specimens. Further study is required to determine whether such analysis may be used in real time to improve surgical decision making and decrease positive surgical margin rates.

Tile-Image Merging and Delivering for Virtual Camera Services on Tiled-Display for Real-Time Remote Collaboration

The tiled-display system has been used as a Computer Supported Cooperative Work (CSCW) environment, in which multiple local (and/or remote) participants cooperate using some shared applications whose outputs are displayed on a large-scale and high-resolution tiled-display, which is controlled by a cluster of PC's, one PC per display. In order to make the collaboration effective, each remote participant should be aware of all CSCW activities on the titled display system in real-time. This paper presents a capturing and delivering mechanism of all activities on titled-display system to remote participants in real-time. In the proposed mechanism, the screen images of all PC's are periodically captured and delivered to the Merging Server that maintains separate buffers to store the captured images from the PCs. The mechanism selects one tile image from each buffer, merges the images to make a screen shot of the whole tiled-display, clips a Region of Interest (ROI), compresses and streams it to remote participants in real-time. A technical challenge in the proposed mechanism is how to select a set of tile images, one from each buffer, for merging so that the tile images displayed at the same time on the tiled-display can be properly merged together. This paper presents three selection algorithms; a sequential selection algorithm, a capturing time based algorithm, and a capturing time and visual consistency based algorithm. It also proposes a mechanism of providing several virtual cameras on tiled-display system to remote participants by concurrently clipping several different ROI's from the same merged tiled-display images, and delivering them after compressing with video encoders requested by the remote participants. By interactively changing and resizing his/her own ROI, a remote participant can check the activities on the tiled-display effectively. Experiments on a 3 × 2 tiled-display system show that the proposed merging algorithm can build a tiled-display image stream synchronously, and the ROI-based clipping and delivering mechanism can provide individual views on the tiled-display system to multiple remote participants in real-time.

Efficient frequency-domain sample selection for recovering limited-support images.

An image whose region of support is smaller than its bounding rectangle can, in principle, be reconstructed from a subset of the Nyquist samples. However, determining such a sampling set that gives a stable reconstruction is a difficult and computationally intensive problem. An algorithm is developed for determining periodic nonuniform sampling patterns that is orders of magnitude faster than existing algorithms. The speedup is achieved by using a sequential selection algorithm and heuristic metrics for the quality of sampling sets that are fast to compute, as opposed to the more rigorous linear algebraic metrics that have been used previously. Simulations show that the sampling sets determined using the new algorithm give image reconstructions that are of accuracy comparable with those determined by other slower algorithms.

Selection of optimal features for classification of electrocardiograms

The forward sequential selection and backward sequential rejection algorithms and the optimal branch and bound algorithm were evaluated in selection of features for classification of electrocardiograms of 237 patients with old myocardial infarction and 299 subjects without infarction. The branch and bound algorithm proved suitable for small sets of ECG features. However, the computational effort required was orders of magnitude greater than that for the other two methods and became prohibitive with large features sets. A satisfactory and consistent overall classification accuracy was achieved by using the sequential selection algorithms for selecting continuous features by maximizing the Mahalanobis distance at each step of the feature selection process. Maximization of the association index can produce better results but requires more computing effort. Feature selection based on maximizing sensitivity at each step for a fixed level of specificity is less satisfactory when a high level of specificity is required.