22 results
Search Results
Now showing 1 - 10 of 22
Item Parallel Particle Swarm Optimization(North Dakota State University, 2016) Manne, PriyankaPSO is a population based evolutionary algorithm and is motivated from the simulation of social behavior, which differs from the natural selection scheme of genetic algorithms. It is an optimization technique based on swarm intelligence, which simulates the bio-inspired behavior. PSO is a popular global search method and the algorithm is being widely used in conjunction with several other algorithms in different fields of study. Modern day computational problems demand highly capable processing machines and improved optimization techniques. Since it is being widely used, it is important to search for ways to speed up the process of PSO, as the complexity of the problems increase. The paper describes a way to improve it via parallelization. The parallel PSO algorithm’s robustness and efficiency is demonstrated. This paper evaluates the parallelized version of the PSO algorithm with the use of Parallel Computing Toolbox available in Matlab.Item Classification Algorithms Applied to a Brain Computer Interface System Based On P300(North Dakota State University, 2017) Gaur, HimanshuA BCI or Brain Computer Interface is defined as a method of communication that converts neural activities generated by brain of living being (without the use of peripheral muscles and nerves) into computer commands or other device commands. BCI systems are useful for people with severe disability who have no reliable control over their muscles in order to interact with their surrounding environment. The BCI system used in this paper has used P300 evoked potential and three classifiers namely Logistic Regression (LR), Neural Network (NN), and Support Vector Machine (SVM). The system is tested with four people with severe disability and two able-bodied people. Classification accuracies obtained from LR, NN, SVM classifiers is then compared with Bayesian Linear Discriminant Analysis (BLDA) classifier and with each other. The relevant factors required for obtaining good classification accuracy in P300 evoked potential based BCI systems is also being explored and discussed.Item Disease Similarity Using Biological Module Dysregulation Profile(North Dakota State University, 2016) Zaman, EshitaDiseases can be grouped according to phenotypic and genotypic similarities. Gene expression and micro-RNA data paved the way to look inside the genetic coding and classify diseases accurately. Modern system biology seeks to understand the underlying protein complexes in a cell and how they are altered in disease condition. In this research, we aimed to mine cohesive biological modules from large micro-RNA dataset and show the genes in these modules are dysregulated in a number of diseases. We used 13 different types of cancer and DME algorithm to extract dense modules satisfying a user defined density. Binary attribute proles of genes are also provided. We have shown that disease similarity based on the average module dysregulation yield disease pairs that share common disease genes. Collectively, we have concluded that the recurrence of these modules in different cancer types increase the therapeutic opportunity to treat more diseases with existing drugs.Item Mining Communities from Multi-Layered Graphs(North Dakota State University, 2013) Chao, MengIdentifying communities from networks has been a subject of great interest in Biological and Social network analysis. Finding communities can help with tasks such as identifying and fighting disease. Using graphs to represent networks and identifying dense subgraphs as communities within these graphs is an increasingly important area of study. Many of the same entities can be found in multiple networks, each representing a different type of relationship. These graphs capturing different relationships between the same entities can be combined into a single graph called a “multi-layered graph”. By finding dense subgraphs containing the same entities within multiple layers of the multi-layered graph, we can increase the confidence these dense subgraphs are communites. This paper has developed an algorithm that takes multi-layered graphs and employs quasi-clique based community discovery for extracting communities. Experimental results on real co-authorship networks show that the proposed approach discover communities that have dense interactions.Item Shortest Path in a Wireless Sensor Network with Multiple Sensor Failures(North Dakota State University, 2011) Poreddy, Sandeep ReddyThis paper shows how a shortest path can be obtained in a wireless sensor network, between a source sensor and a destination sensor, in a hop-by-hop fashion, considering multiple sensor failures along the path of data transmission. A wireless sensor network consists of a number of sensors spread across a geographical area; these sensors are distributed either randomly or systematically. Every sensor possesses the capability to communicate with other sensors, and each sensor possesses some level of intelligence to perform processing of the signals. Efficiency of a sensor network depends on the design of the network topology. The network must be continuously functional to perform the task; a major problem in wireless sensor networks is due to sensor failures. In order to remain functional in spite of sensor failures, it is required that an alternative shortest path is found to send and receive requests to fulfill the task. I addressed this problem by finding an alternative shortest path based on the hop count. I developed a web based application to simulate a network and find shortest paths in a network with multiple sensor failures. I performed an experimental analysis in finding the shortest path. When a source sensor has data to transmit to a destination sensor, it broadcasts a RREQ (Route Request) to its immediate neighbors. A route to the source is created at every sensor when a RREQ is received. If the receiving sensor has not received this RREQ before and if it is not the destination, then it broadcasts the RREQ to its immediate neighboring sensors. If the receiving sensor is the destination, it generates a RREP (Request Reply). The RREP is uni-cast to the source sensor in a hop-by-hop fashion and a shortest path is obtained. Broadcasting is stopped at the point where a destination sensor receives a RREQ and acknowledges by RREP. To obtain a new shortest path, one or more sensors in the previously obtained shortest path are failed and the algorithm continues from the point where it has broadcasted RREQ to the sensors in the network previously and a new shortest path is obtained.Item Performance Comparison of Apache Spark MLlib(North Dakota State University, 2018) Sharma, PallaviThis study makes an attempt to understand the performance of Apache Spark and the MLlib platform. To this end, the cluster computing system of Apache Spark is set up and five supervised machine learning algorithms (Naïve-Bayes, Decision Tree, Random Forest, Support Vector Machine and Logistic Regression) were investigated. Among the available cluster modes, these algorithms were implemented on two cluster modes, Local and GPU Cluster mode. The performance metrics such as classification accuracy, area under ROC and area under PR for the algorithms were investigated by considering three datasets. It is concluded that the algorithms are computed in parallel in both the modes with GPU Cluster mode performing better than the Local mode for all algorithms in terms of time taken for completion. However, the mentioned performance metrics were not affected in the two modes hinting that the parallel computation does not play a major role in determining these metrics.Item Design and Evaluation of Two Hybrid Genome Assembly Approaches Using Illumina, Roche 454, and PacBio Datasets(North Dakota State University, 2016) Sun, LirenThe assembly of next-generation sequencing reads is one of the most challenging and important tasks in bioinformatics. There are many different types of assembly algorithms and programs that have been developed to assemble next-generation sequencing reads. However, the assembly quality of each assembly program may vary. This paper introduces and implements two different assembly approaches that use three types of next-generation sequencing datasets. Both assembly approaches are designed to achieve the same goal, which is to improve assembly quality. The assembly results from the two approaches were compared and evaluated by using some widely used quality metrics. The result shows each approach has advantages and disadvantages.Item Algorithms for Coverage Improvement in a Sensor Network(North Dakota State University, 2011) Maddi, Sunil ReddySensors are devices which have the ability to receive and respond to a signal. These sensors, when used as a group, form a sensor network. Sensors in a sensor network can communicate and transmit data. In the early stages of research on sensor networks, only static sensors were used to form a sensor network. As research advanced, a combination of static and mobile sensors was used to form a wireless sensor network instead of just static sensors. The primary advantage of this type of sensor network over a sensor network with static sensors is the ability of mobile sensors to move to a new location in the network to increase the overall area covered by the sensors. Some concerns in a wireless sensor network are coverage area, energy consumption of the sensors, the ratio of static and mobile sensors to be used in sensor network, and the deployment of sensors in a network. Major research in sensor networks is focused on addressing the issue of coverage area. The objective of this paper is to design, implement and analyze Most Overlapped First and Highest Coverage Gain algorithms that address the issue of coverage area in a wireless sensor network. Local Spiral Search was used as the base to develop these two algorithms in combination with the Utility Function. Both algorithms were tested, and the results were analyzed with coverage area and change in overlap area as the metrics. Results showed a significant gain in coverage area using both the algorithms, and there was a consistent change in overlap area for a varying number of sensors.Item A Network Optimization Solver for Routing in Wireless Sensor Networks(North Dakota State University, 2010) Samaraweera, Shaminda AselaMany wireless sensor network applications require energy efficient communication between nodes in the network. Sensor networks are of limited resources. Due to this limitation, the routing between the nodes is one of the important aspects of the life span of the total network. Optimization of the routing algorithm is therefore an important decision point in the design of the sensor network. Our study establishes that optimization can increase the life span of the network. We implement an optimization algorithm in the total network, which is capable of saving energy on communication. The energy saving in communication helped us to increase the life span of the network.Item Automated Detection of Acute Leukemia Using K-Means Clustering Algorithm(North Dakota State University, 2019) Arya, MinakshiDetection of ALL can be done through the analysis of white blood cells (WBCs) called leukocytes. Usually, the analysis of blood cells is performed manually by skilled operators, have numerous drawbacks, such as slow analysis, a non-standard accuracy and skill of the operator. Hence many automated systems are using in order to analyze and classify the blood cells. This paper focuses on an automatic system based on image processing algorithms for the classification of blood cells for detection of Acute Lymphocytic Leukemia (ALL). Experiments were ran using 20 models with PCA and seven models namely Medium KNN, Coarse KNN, Cosine KNN, Cubic KNN, Weighted KNN, Ensemble Boosted trees and Ensemble Bagged trees had 99.9% accuracy. These models are evaluated based on the prediction speed, training time, confusion matrix and ROC. Of all models, the weighted KNN classifier is best when using PCA.
- «
- 1 (current)
- 2
- 3
- »