Department: Engineering and Applied Science: Computer Science ![[clear]](close-x.png "Remove this limiter")
23 matches in the database.
These are records: 1 - 23.
Did you mean instcode:ucii?
1.
Carraher, Lee A.
A Parallel Algorithm for Query Adaptive, Locality Sensitive Hash Search.
Degree: MS, Engineering and Applied Science: Computer Science, 2012, University of Cincinnati
► Nearest neighbor search is a fundamental requirement of many machine learning algorithms…
(more)
▼ Nearest neighbor search is a fundamental requirement of many machine learning algorithms and is essential to fuzzy information retrieval. The utility of efficient database search and construction has broad utility in a variety of computing fields. Applications such as coding theory and compression for electronic communication systems as well as use in artificial intelligence for pattern and object recognition. In this thesis, a particular subset of nearest neighbors is consider, referred to as c-approximate k-nearest neighbors search. This particular variation relaxes the constraints of exact nearest neighbors by introducing a probability of finding the correct nearest neighbor c, which offers considerable advantages to the computational complexity of the search algorithm and the database overhead requirements. Furthermore, it extends the original nearest neighbors algorithm by returning a set of k candidate nearest neighbors, from which expert or exact distance calculations can be considered. Furthermore thesis extends the implementation of c-approximate k-nearest neighbors search so that it is able to utilize the burgeoning GPGPU computing field. The specific form of c-approximate k-nearest neighbors search implemented is based on the locality sensitive hash search from the E2LSH package of Indyk and Andoni [1]. In this paper, the authors utilize the exceptional properties of the Leech Lattice [2], as a subspace quantizer for the locality sensitive hash families. The Leech Lattice is unique in that it provides the closest lattice packing of equal sized spheres in 24 dimensional space. In addition, advances from coding theory provide a very favorable decoding algorithm for finding the nearest lattice center to a query point in euclidean 24 dimensional space [3] [4]. The multilevel construction of the Leech Lattice provides an excellent opportunity for parallelization as it contains the minimization of many independent sub-lattice decodings resulting from the lattices exceptional symmetry among lattices. These decodings are additionally highly floating point computationally intensive, and because of which suggest a favorable implementation on GPGPU architectures such as NVIDIAs CUDA based framework. Furthermore, the overall construction of a locality sensitive hash based, nearest neighbors search algorithm, is able to be parallelized fairly efficiently as the hash decodings are completely independent of one another. The goal of this thesis is to present a CUDA optimized parallel implementation of a bounded distance Leech Lattice decoder [4] for use in query optimized c-approximate k-nearest neighbors using the locality sensitive hash framework of E2LSH. The system will be applied to the approximate image retrieval of SIFT transformed [5] image vectors. [1] A. Andoni, Nearest Neighbor Search: the Old, the New, and the Impossible. INSTITUTE OF [2] J. Leech, “Notes on sphere packings,” [3] . J. Forney, G. D., “A bounded-distance decoding algorithm for the leech lattice, with generalizations,” [4] O. Amrani and Y. Beery, “Efficient bounded-distance decoding of the hexacode and associated decoders for the leech lattice and the golay code,” [5] D. G. Lowe, “Object recognition from local scale-invariant features,”
Advisors/Committee Members: Annexation, Fred.
Subjects: Computer Science
Keywords: Locality Sensitive Hashing; Approximate Nearest Neighbors; CUDA GPU; Image Search; Distance Adaptive LSH; Parallel Computing
More Like This
2.
Chase, Chelsea.
Using Design Patterns in User Interface Design.
Degree: MS, Engineering and Applied Science: Computer Science, 2012, University of Cincinnati
► User interface development is a large part of software development and is…
(more)
▼ User interface development is a large part of software development and is highly visible to the user. It’s important to create a good user interface but it can be very time consuming. The use of design patterns can be beneficial to speed up user interface design, as well as implementation of common features. Design patterns are high-level solutions to common problems in software development. These patterns have been widely used in object-oriented programming and across many different programming languages. Each design pattern has a name and a set of attributes that make up the pattern. So far, no standard set of attributes has been agreed upon. We present an easy, all-inclusive list of attributes specifically developed for use with interaction patterns, the special type of design pattern used in user interface design. This list helps to bridge the gap left by other proposed attribute lists and we use it to define several design patterns identified in our case study of Facebook and its mobile app for iPhone and Android devices.
Advisors/Committee Members: Annexstein, Fred.
Subjects: Computer Science
Keywords: User Interface; UI; Interaction; Design; Pattern; Facebook
More Like This
3.
Dalal, Avani.
Interference Analysis and Mitigation in a Cellular Network with Femtocells.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Development of femtocells is the most recent step towards increasing the network…
(more)
▼ Development of femtocells is the most recent step towards increasing the network capacity of a wireless network and improving the quality of service for cellular users. However, large scale femtocell deployment can significantly affect existing wireless network. We examine the effect of interference due to closed-access femtocell deployment in a wireless network on macro users. Numerical analysis and simulation are performed in order to understand the impact of femtocells in an existing wireless network. Another interference scenario arising in such networks is the interference amongst neighboring femtocells. This is due to the fact that all the femtocells use the same radio band, and thus it poses a serious interference issue amongst neighboring femtocells, especially when femtocells are densely deployed. We study the problem of downlink and uplink interference used in conjunction with the LTE technology and propose a solution to mitigate this interference using Fractional Frequency Reuse (FFR) technique. User equipments (UEs) report to their serving femto base stations (F-BS) periodically. If the report reveals high interference level, neighboring F-BSs locally adopt FFR in a coordinated but decentralized way. FFR divides the femtocell in cell-center and cell-edge areas where the cell-edge of interfering F-BSs use distinct set of sub carriers. The division takes place in a way such that the affected UE falls in the cell-edge area. The simulation results indicate that the proposed scheme decreases interference in the neighboring F-BSs, femto UEs (F-UEs) and macro UEs (M-UEs) and improves the overall throughput.
Advisors/Committee Members: Agrawal, Dharma.
Subjects: Computer Science
Keywords: Femtocell; Macrocell; Fractional Frequency Reuse; Reverse Gaussian Distribution; Macro user; Interference
More Like This
4.
Gaur, Amit.
Secured Communication in Wireless Sensor Network (WSN) and Authentic Associations in Wireless Mesh Networks.
Degree: MS, Engineering and Applied Science: Computer Science, 2010, University of Cincinnati
► Wireless sensors are low power devices with small transmission range, restricted computation…
(more)
▼ Wireless sensors are low power devices with small transmission range, restricted computation power, limited amount of memory and with portable power supply. Wireless Sensor Network (WSN) is a collection of such sensors where the number of sensors can vary from few hundreds to thousands. Performing secure pair-wise communication between sensors is a really difficult task due to inherent characteristics such as lack of any fixed infrastructure. As memory and power consumption are most stringent requirements for these devices, use of conventional techniques for secured communication are totally out of question. This thesis introduces scheme that enables a complete pair-wise secure connectivity between any two adjacent sensor nodes in spite of using small key ring (KR) for sensors. The Proposed Scheme (ELKPD) doesn't require any additional hardware while providing keys to the sensors irrespective of their location. Also, proposed scheme is easily scalable which allows enables addition of sensor nodes without any computational or hardware overheads. Due to the varying degree of mobility of Mesh Clients has provided much more flexibility in Wireless Mesh Networks. And establishing an Authentic Association among entities is a non -trivial problem. In this thesis, we introduce a Polynomial Based scheme which not only provides high pair-wise connectivity, low communication and storage overhead and high scalability but also makes on the fly Authentic Association feasible. The proposed scheme is also observed to be resilient against both the traffic analysis and the node capture attacks.
Advisors/Committee Members: Agrawal, Dharma.
Subjects: Computer science
Keywords: security; wireless sensor networks; key-predistribution; wireless mesh networks; bi-variate polynomials
More Like This
5.
Gupta, Nidhi.
Mutual k Nearest Neighbor based Classifier.
Degree: MS, Engineering and Applied Science: Computer Science, 2010, University of Cincinnati
► In this information intensive world, data is used to make statistical decisions…
(more)
▼ In this information intensive world, data is used to make statistical decisions for business, scientific, and industrial situations. With decreasing cost of computing power and storage, data with various properties are available to data miners for classification type of decisions. So far the classifiers designed to classify the datasets with varying data density required too many parameters and the methods did not work very well. In this research we have introduced a new idea called the Mutual k-Nearest Neighbor (Mk-NN) relationship and designed classifiers based on this idea. We have validated our results with a number of synthetic and a real-world data set and have received superior results. We present these result of the classifications in this thesis and also describe our approach and algorithms.
Advisors/Committee Members: Bhatnagar, Raj.
Subjects: Artificial Intelligence
Keywords: K Nearest Neighbor; Mutual k Nearest Neighbor; Classifier; Classification algorithm; Reverse Nearest Neighbor
More Like This
6.
Hausrath, Nathaniel L.
Methods for Hospital Network and Computer Security.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Hospital IT security presents many unique challenges that must be solved by…
(more)
▼ Hospital IT security presents many unique challenges that must be solved by the entire organization. Network and computer threats can cause thousands of dollars in lost time and resources, legal repercussions, and damaged repu- tation. Despite warnings from a wealth of public breach notifications, many hospitals are inadequately prepared to deal with today’s computer-based at- tacks. This thesis explores the root causes of hospital network and computer in- security, and addresses these problems with methods implemented in actual hospitals. A lack of comprehension of methods to assess and implement secu- rity measures by hospital IT security employees can hinder network visibility and prevent their ability to stop threats. In addition, these same people are unable to express security concerns in terms management can understand, harming their credibility within the business as a whole. Without this sup- port, organizational change is impossible. By addressing these concerns with a combination of people, process, and tools, we can solve complex problems, protect patient data, and ensure IT operations so hospitals can serve their community and save lives.
Advisors/Committee Members: Franco, John.
Subjects: Information Technology
Keywords: hospital it Security; information security; network security; computer security; hospital information security; security
More Like This
7.
Joshi, Saugat.
Fast and Efficient Mutual Authentication in Wireless Mesh Networks (WMNs).
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Through the evolvement of high speed internet, Wireless mesh networks (WMNs) have…
(more)
▼ Through the evolvement of high speed internet, Wireless mesh networks (WMNs) have become one of the most exciting and promising technology for providing high bandwidth features to its users. Along with the advancement of internet, the demand for enhanced capacity and higher bandwidth requirement have strived over time to meet the requirements of the Quality of Service (QoS) in WMNs. Various factors do affect the desired Qos for WMNs. In this thesis, we focus on the key generation scheme for authentication between various entities of the network to establish a secure communication between them, while taking into consideration the QoS requirements set by the benchmark. The key generation scheme discussed here is decentralized and hierarchal in nature which enables a pair of entity (e.g., servers and clients) to share a common key for a secure communication. Moreover, the scheme addresses the issue of the high speed mobility of the clients stations (STAs) from one domain to another domain, i.e. handoff between various inter- domain and intra- domain Access Points (APs). It is necessary that the STAs do not require excessive overhead during the handoff procedure. The schemes discussed enables faster and secure key generation and agreement scheme between the entities of the network during the handoff procedure. The key generation scheme is distributed in nature. The higher level hierarchy namely the Internet Gateways (IGWs) or the authentication, authorization and accounting (AAA) servers such as RADIUS, generate a multi variate symmetric polynomial function and exchange the information among them such that none of them have a complete knowledge of the entire generated function. As the functions are passed to the lower level hierarchical entities such as Mesh access points (MAPs), the function further reduces providing only legitimate information to them. The process continues until the lowest level of the hierarchy (STAs or clients) is reached where the entities will be able to deduce a secure key for the communication. We refer to this as a distributed mechanism or a distributed authenticated key establishment (AKE) scheme based on hierarchal multi-variable symmetric functions (HMSF). Since, the deduced key is obtained from distributed scheme and below various levels none of the entities have a complete knowledge to reverse engineer the original function used in the generation process. Using the distributed authenticated key establishment scheme the STAs and MAPs could authenticate among themselves without any assistance from the higher hierarchy entities, thus saving the communication overhead time and the delay involved in authentication by getting back to the servers hence maintaining the required QoS.
Advisors/Committee Members: Agrawal, Dharma.
Subjects: Computer Science
Keywords: Wireless Mesh Networks; Authentication; symmetric; polynomial; handoff; security
More Like This
8.
Kaur, Jasman.
Realizing Connectivity with Independent Trees in DAGs - An Empirical Study.
Degree: MS, Engineering and Applied Science: Computer Science, 2012, University of Cincinnati
► Reliability and fault tolerance are an important part of data gathering at…
(more)
▼ Reliability and fault tolerance are an important part of data gathering at a base station (sink) in a wireless sensor network. Due to the uncertain environments of the network it becomes necessary to have robustness and information security integrated into the routing schemes. Multipath routing using independent trees is an eective way to achieve it. In real world applications, every node of the network does not have homogeneous connectivity to the base station. A more practical approach and a recent research area is having heterogeneous node connectivity with independent trees which are not necessarily spanning. A recent algorithm, referred to as DAG Independent Trees that gives vertex independent trees in heterogeneously connected directed acyclic graphs (DAGs) is discussed. The problem of realizing the vertex connectivity in independent trees is NP-complete. Therefore, an experimental analysis of the connectivity realized with independent trees given by the algorithm, and vertex connectivity from the nodes to sink is given. The results show that the algorithm nearly satises the condition that all nodes realize at least half the connectivity. A close approximation to the connectivity at more than half the nodes is achieved. In the case, when the algorithm does less than half the connectivity for most of the nodes, an optimization of the algorithm is given. The optimization over all permutations of the sink nodes is presented in the case when it does not fair well.
Advisors/Committee Members: Berman, Kenneth.
Subjects: Computer Science
Keywords: Connectivity; Independent trees; DAGs; Multipath Routing; Data gathering; Wireless Sensor Networks
More Like This
9.
Kushwaha, Akash.
Methods for estimation of voters' weights for weighted rank aggregation.
Degree: MS, Engineering and Applied Science: Computer Science, 2012, University of Cincinnati
► Rank aggregation is the problem of combining ranking of candidates (alternatives, players,…
(more)
▼ Rank aggregation is the problem of combining ranking of candidates (alternatives, players, products) from different sources (voters, critics, judges, product reviewers) to obtain a consensus ranking. Researchers have used rank aggregation methods for problems in sports, information retrieval, machine learning and computational biology. There are aggregation problems where accurate aggregation requires using weights for voters, e.g. , experts of a field are known to have varying performance. While aggregating their opinion on a topic, it is desired to weight opinion of an expert differently. External validation measures have been used as weights to perform weighted rank aggregation. However, there are no unsupervised methods available which estimate weights from given data. In this work, we propose two methods for estimating voters weights from the given data. We also describe an algorithmic framework that allows coupling of any rank aggregation algorithm with proposed methods to perform weighted rank aggregation. This framework is an iterative method which generates weight vector and weighted ranking of candidates. Convergence of framework is a necessary condition for obtaining voters weights. We experimented with three publicly available real datasets and randomly generated datasets. Experiments provide empirical proof of convergence of proposed methods.
Advisors/Committee Members: Berman, Kenneth.
Subjects: Computer Science
Keywords: weighted rank aggregation; opinion spam; ranking; unsupervised methods
More Like This
10.
Lund, Benjamin.
Some Results in Discrete Geometry.
Degree: MS, Engineering and Applied Science: Computer Science, 2012, University of Cincinnati
► This thesis includes several results on the extremal combinatorics of hyperplane arrangements,…
(more)
▼ This thesis includes several results on the extremal combinatorics of hyperplane arrangements, kinetic point sets, and other geometric struc- tures. It consists of three papers published or submitted elsewhere. “A Bichromatic Incidence Bound and an Application”, joint with George Purdy and Justin Smith, was published in Discrete and Computational Geometry. I presented “Collinearities in Kinetic Point Sets”, joint with George Purdy, Justin Smith, and Csaba T´th, at the 23rd Canadian Con- o ference on Computational Geometry, 2011. “Some Results Related to a Conjecture of Dirac’s”, joint with George Purdy and Justin Smith, has been submitted to The Electronic Journal of Combinatorics.
Advisors/Committee Members: Purdy, George.
Subjects: Computer Science
Keywords: combinatorial geometry; hyperplane arrangements; pseudoline arrangements; extremal combinatorics; kinetic points
More Like This
11.
Mahadevan, Srisudha.
Network Selection Algorithm for Satisfying Multiple User Constraints Under Uncertainty in a Heterogeneous Wireless Scenario.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► The constant evolution of various wireless access technologies has led to an…
(more)
▼ The constant evolution of various wireless access technologies has led to an advancement in the communication domain where mobile clients (MCs) are equipped with multiple interfaces for simultaneous access to different types of networks. This heterogeneous wireless scenario satisfies the user's preferences and offers the desired quality of service (QoS). Route selection satisfying multiple constraints has proven to be NP-Complete and various approximation schemes exist which consider the network state resources to be fixed and complete. Additionally, in practical scenarios, the user's constraints are imprecise and vague and the changing network conditions force the condition of uncertainty to exist in a routing mechanism. This thesis focuses on the imprecise and dynamic nature of network parameters and maps it to uncertain user constraints. We, hence consider a heterogeneous wireless network (HWN) and propose a novel approach to identify the key network metrics that satisfies the user's criteria. Our design of a fuzzy model maps the underlying uncertainty in the metrics to crisp values and demonstrates the stability of our proposed technique through extensive simulations and analysis. In order to satisfy the user's imprecise demands, we consider the problem of decision making in a HWN where emphasis is on selecting the best network interface to forward the data. We propose an enhanced minimization of maximal regret (MMR) approach to rank the available network interfaces by considering pure uncertainty in a user's constraint. MCs state their needs based on application requirements and thus, we implement a generalized MMR and include Ordered Weighted Averaging (OWA) operators to enable each MC to efficiently select the best possible alternative. The weights utilized in OWA are modeled using application characteristics. Our simulations and experiments compare the sensitivity of user demands depicted in MMR and OWA to that of existing multiple attribute decision making (MADM) algorithms. Simulation results prove that variability in a user's preference and changes in a network scenario can impact decision making and influence the routing process.
Advisors/Committee Members: Agrawal, Dharma.
Subjects: Computer Science
Keywords: Network Selection; Heterogeneous Wireless Network; Fuzzy Logic; Decision Making; Multiple metrics; Uncertainty
More Like This
12.
Narayanan, Sriram.
Proportional Fairness in Regular Topologies of Wireless Sensor Networks.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Wireless Sensor Networks (WSNs) are being used in a wide variety of…
(more)
▼ Wireless Sensor Networks (WSNs) are being used in a wide variety of commercial civilian applications which favour regular deployment of sensor nodes (SNs) over a random deployment due to its simplicity of planning, ease of maintenance and precise node location information. In such regular deployments, it is necessary to ensure that every SN is allocated a fair share of the network’s resources in order to successfully obtain the sensed information from every portion of the area covered by the WSN. However, in the interest of maintaining the best possible network performance, the total throughput of the WSN needs to be maximized. Thus, we consider the problem of proportionally fair rate allocation to the SNs in regular WSNs to achieve maximum throughput while ensuring that every SN is allocated a fair portion of the available bandwidth. We follow a cross layer approach, considering both the transport layer session rates and the link layer transmission probabilities required to achieve the maximum total proportionally fair throughput for the square, hexagonal, and triangular topologies of WSN. Closed form expressions for each topology are derived, and they enable us to determine which topology has the best performance at different network sizes. Network simulations conducted in Qualnet and experiments conducted using TELOSB SNs validate our theoretical results and shed light on future directions of work.
Advisors/Committee Members: Agrawal, Dharma.
Subjects: Computer Science
Keywords: Wireless Sensor Networks; Regular Topologies; Proportional Fairness
More Like This
13.
Nepal, Srijan.
Linguistic Approach to Information Extraction and Sentiment Analysis on Twitter.
Degree: MS, Engineering and Applied Science: Computer Science, 2012, University of Cincinnati
► Social media sites are one of the most popular destinations in today’s…
(more)
▼ Social media sites are one of the most popular destinations in today’s online world. With millions of users visiting social networking sites like Facebook, YouTube, Twitter etc. every day to share social content at their disposal; from simple textual information about what they are doing at any moment of time, to opinions regarding products, people, events, movies to videos and music, these sites have become massive sources of user generated content. In this work we focus on one such social networking site - Twitter, for the task of information extraction and sentiment analysis. This work presents a linguistic framework that first performs syntactic normalization of tweets on top of traditional data cleaning, extracts assertions from each tweet in the form of binary relations, and creates a contextualized knowledge base (KB). We then present a Language Model (LM) based classifier trained on a small set of manually tagged corpus, to perform sentence level sentiment analysis on the collected assertions to eventually create a KB that is backed by sentiment values. We use this approach to implement a contextualized sentiment based yes/no question answering system.
Advisors/Committee Members: Berman, Kenneth.
Subjects: Computer Science
Keywords: Sentiment Analysis; Twitter; Information Extraction; Language Model; Machine Learning; Sentiment Classification
More Like This
14.
Nerusupalli, Sathvik.
Personalized User Trending Topics.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Online Social networks have a cornucopia of interesting information waiting to be…
(more)
▼ Online Social networks have a cornucopia of interesting information waiting to be tapped and understood. Trending topics is one of the easier ways to understand the inclination of information that is being published in social networks. Trending topics are generic and might not be of interest to a certain individual on every occasion. Personalized user trends are the trending topics that occur in the neighborhood of a user. Further, in this thesis, we discuss techniques to find personalized user trends by using a new process in stop word removal that is effective in dynamic environment like Twitter and in identifying features that affect trending topics in a user’s neighborhood.
Advisors/Committee Members: Annexstein, Fred.
Subjects: Computer Science
Keywords: Twitter; Trending Topics; Personalized; Trends; stop words
More Like This
15.
Shahdeo, Sandhya.
MiR-Drug Relationships: Mining and discovering bi-domain dense subclusters using greedy randomized algorithm.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Gene regulation, primarily achieved at the transcriptional level, is central to the…
(more)
▼ Gene regulation, primarily achieved at the transcriptional level, is central to the normal development and functioning of all organisms. It therefore represents an obvious target for therapeutic drugs which could act either by stimulating or inhibiting specific gene transcription to elicit desired effects. Likewise, microRNAs, the recently discovered small non-coding RNAs, negatively regulate target genes post-transcriptionally through degradation or suppression of protein translation. Since both these regulators, endogenous and exogenous, affect biological and chemical pathways targeting gene regulation, the question is whether they can be used alternatively or combinatorially. Thus, hypothesizing that microRNAs- the endogenous regulators (mostly suppressors) of genes, could be used as alternatives to drugs or as combinatorials (with drugs) to fine tune the drug-response or mitigate potential side-effects, we used existing techniques like clustering, self organizing maps and computation of nearness in n dimensional space, to analyze the correlation between drugs and microRNAs. An algorithm to incorporate the heterogeneity of data by introducing a per data-species ratio, using a randomized greedy approach, was conceived. This algorithm finds the largest square matrices complying with the drug/microRNA ratio and density threshold by randomly selecting seed matrices and then systematically growing the best ones available and converges over multiple runs.
Advisors/Committee Members: Cheng, Yizong.
Subjects: Computer Science
Keywords: microRNA drug relation; greedy randomized algorithm
More Like This
16.
Sharma, Abhishek D.
Discovery of Trajectory Clusters in Spatio-Temporal Data.
Degree: MS, Engineering and Applied Science: Computer Science, 2010, University of Cincinnati
► Extracting and Clustering of trajectories from Spatio-Temporal data is a challenging problem…
(more)
▼ Extracting and Clustering of trajectories from Spatio-Temporal data is a challenging problem due to the highly exponential nature of the space of possible clusters. It is an important problem as it holds potential for obtaining insights into vast amounts of spatio-temporal data. Most of the existing algorithms for mining spatio-temporal data focus on clustering the sets of meaningful trajectories that are already identified and available to them. Discovering interesting trajectory clusters in a three dimensional spatio-temporal dataset in which each cell has been instantiated to a value is an extremely large problem due to the number of potential trajectories. In this thesis we present an algorithm which uses divide-and-conquer strategy by finding clusters in few layers at a time and then combining these results to construct larger clusters. We demonstrate how this strategy emulates results that may be obtained for the complete datasets.
Advisors/Committee Members: Bhatnagar, Raj.
Subjects: Computer science
Keywords: Spatio Temporal
More Like This
17.
Smith, Justin Wesley.
Points and Lines in the Plane.
Degree: MS, Engineering and Applied Science: Computer Science, 2010, University of Cincinnati
► This thesis will focus on two topics: (1) finding intersections determined by…
(more)
▼ This thesis will focus on two topics: (1) finding intersections determined by an arrangement of hyperplanes (e.g., lines in a plane), and (2) lower bounds on the number of various “types” of lines determined by a configuration of points. The first topic is algorthmic. Given an arrangement of n lines in ℝ2, a O(n log n) algorithm is demonstrated for finding an ordinary intersection (i.e., an intersection of exactly two lines). This algorithm is then extended to finding an ordinary intersection among hyperplanes in ℝd, under the hypothesis that no d hyperplanes pass through a line and not all pass through the same point. Algorithms are also given to nd an ordinary intersection in an arrangement of pseudolines in time O(n2), and to find a monochromatic intersection in a bichromatic arrangement of pseudolines in time O(n2). The second topic is combinatorial. Let G and R be finite sets of points, colored green and red respectively, such that |G| = n, |R| = n−k, G ∩ R = 0, and G ∪ R are not all collinear. Lower bounds will be demonstrated for several types of lines (e.g., bichromatic and equichromatic) determined by few points in ℝ2.
Advisors/Committee Members: Purdy, George.
Subjects: Computer science
Keywords: bichromatic; points; lines; geometry; discrete; combinatorics
More Like This
18.
Subramanian, Hema.
Summarization Of Real Valued Biclusters.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► With an explosion in database sizes, there is an increasing need for…
(more)
▼ With an explosion in database sizes, there is an increasing need for mining relevant information from them. Subspace clustering has been applied in various fields for discovering patterns, and many such algorithms have been investigated for finding interesting biclusters from binary-valued datasets. Mining biclusters from real-valued datasets has gained significant importance in many of the recently emerging applications. The algorithms devised for mining such biclusters generally minimize an objective function, and hence the biclusters generated by each algorithm vary depending on the objective function used. Due to the inherent size and density of the data sets, the algorithms generate a very large number of biclusters, making it dicult to select the useful ones from among them. To overcome this problem, it is important to design strategies to summarize these biclusters into few representatives of the main ideas embedded in the dataset. The objective of this thesis is to apply some statistical properties of the generated biclusters to identify some distinguished clusters that seek to summarize the large number of biclusters into few representative ones. In order to achieve the above stated objective, similarity measures based on mutual information and standard deviation d between biclusters are used to identify similar biclusters. These measures quantify the information shared (or the similarity) between two biclusters, and this helps in identifying potential biclusters that could be merged. The algorithm has been applied to a synthetic and two real world datasets and the results are presented. The information content and the variance in a bicluster are analyzed as the biclusters are progressively merged. The methodologies proposed in this thesis are compared to a baseline method to verify the quality of the biclusters and validate that our approach performs significantly well and has good merit.
Advisors/Committee Members: Bhatnagar, Raj.
Subjects: Computer Science
Keywords: biclusters; real-valued biclusters; bicluster reduction; formal concept analysis
More Like This
19.
Sugandharaju, Ravi Kumar Chatnahalli.
Gaussian Deconvolution and MapReduce Approach for Chipseq Analysis.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Protein DNA interactions are one of the most fundamental factors used to…
(more)
▼ Protein DNA interactions are one of the most fundamental factors used to determine how proteins and transcription factor binding sites affect the mechanisms related to a phenotype. The path breaking methodology CHIP-Seq which is a result of the combination of Chromatin Immunoprecipitation with parallel DNA sequencing in identifying the above interactions are of great interest. A methodology to identify the underlying peaks which are a direct representation of the protein interactions with the DNA is of paramount importance in identifying genes and pathways. We propose GaussDeconvo, a novel Gaussian deconvolution approach, which is very precise in identifying these peaks, separating closely spaced peaks and it is also one of the first steps towards the comparative analysis of CHIP-Seq drawn from many experiments. We also propose a scaling mechanism for our computationally intensive algorithm, by successfully mapping the problem to a functional programming domain and applying Hadoop MapReduce to process the data in a distributed setup. We also show that the algorithm scales linearly with the increase of nodes in the cluster.
Advisors/Committee Members: Bhatnagar, Raj.
Subjects: Computer Science
Keywords: dna sequencing; chipseq; mapreduce; hadoop; deconvolution; nedler mead
More Like This
20.
Tennety, Srinivas.
Mobile robot navigation in hilly terrains.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Mobile robot navigation in hilly terrains is challenging since the environment is…
(more)
▼ Mobile robot navigation in hilly terrains is challenging since the environment is unstructured, ill-conditioned and complex. The features of the terrain cannot be easily classified as traversable or non-traversable and therefore identification of paths that pose minimum danger to the robot becomes difficult. One approach to navigation in hilly terrain is based on unsupervised learning where robot learns via interacting with the environment based on trial and error. This method can be implemented using reinforcement learning. However, this approach is not applicable to real world applications as the robot might incur unrecoverable damage while interacting with the environment. Another approach is using human expert knowledge. Humans learn from their past experiences and display an uncanny ability to identify safe paths even in the presence of uncertainties. Therefore, it is beneficial to use the human expert knowledge when available, to aid in navigation of robots in complex terrains. This thesis presents a framework where human expert assistance is used to guide the robot to the goal through reinforcement learning techniques. When a prior knowledge of the terrain such as low resolution aerial view is available, a human expert can identify one or more paths from start to goal that are relatively safe to traverse. These expert paths are used to approximate a value matrix that would steer the robot from any start position in the terrain to the goal avoiding any unsafe regions that pose obvious danger. This approach aids in global path planning and does not take local terrain information in to the consideration that might not be available to the expert. To facilitate incorporation of local terrain information, a fuzzy logic controller is designed which can be used to update the value matrix based on the local sensor data. Experiments have been carried out in simulated hilly terrains with and without the expert paths to show the effectiveness of the approach. Different scenarios have been discussed to demonstrate the advantages of specifying multiple expert paths over few and also the integration of the fuzzy logic controller.
Advisors/Committee Members: Bhatnagar, Raj.
Subjects: Robots
Keywords: Mobile robot; Reinforcement learning; Hilly terrains; Autonomous Navigation
More Like This
21.
Venkataraman, Aparna.
Dynamic Deployment strategies in Ad-Hoc Sensor networks to optimize Coverage and Connectivity in Unknown Event Boundary detection.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► There are many ways to geographically determine the boundary of an event…
(more)
▼ There are many ways to geographically determine the boundary of an event based on its location and its nature through Satellite imaging and other learning mechanisms. In these methods, the availability of resources to perform the detection, their capabilities, actual time available to determine the event and its accessibility are constraints. At times, the satellite images may not be sufficient to get complete information about an event. Here we consider a particular case where the aim is to detect the actual boundary of an event based on its estimated boundary with the above constraints. A typical situation would be to determine the actual boundary of fire given the smoke area, or to estimate the concentration of chemical content, ideally any situation where sensors need to be used in an unmanned situation. We use a deploying agent to drop the sensors and there is a Base Station (BS) to which the event details are communicated by connectivity through localization with neighboring sensors. The research targets dynamic deployment of sensors with coverage and connectivity handled simultaneously as the information can reach the base station only if the sensors are able to connect to it. This is critical for real time applications. So we use an intelligent distribution scheme to test the behavior of different kinds of deployments using random, Gaussian, controlled random and combinational methods to deploy sensors. The set of parameters which are constraints are the communication radius of the base station, sensors & the event, the proximity of the event from the base station and location determination of the event based on the current state of the system. We use a weighted approach with more sensors around the event border and lesser inside to be able to detect the event and yet preserve the sensors as they might be lost due to fire or damage depending on the event. Additionally partial event boundary detection is used as experiment results show that we can reduce the number of sensors by nearly 50% when 80% of the sensors deployed are connected as against 100% connectivity. Test cases also involve multiple BS & deploying agents with parallel control. This will be useful in emergency situations and specifically in situations which do not have pre-deployed sensors wherein, time and resources available are important constraints.
Advisors/Committee Members: Agrawal, Dharma.
Subjects: Computer Science
Keywords: Dynamic network deployment strategies; realtime partial boundary detection; ad-hoc wireless sensor networks; localization; coverage connectivity detectivity; unknown event boundary detection
More Like This
22.
Venkatesan, Vaidehi.
Cuisines as Complex Networks.
Degree: MS, Engineering and Applied Science: Computer Science, 2011, University of Cincinnati
► Cuisines are among the richest and most distinctive artifacts of human culture,…
(more)
▼ Cuisines are among the richest and most distinctive artifacts of human culture, reflecting social structures, religious constraints, ecological choices, economic factors, historical events and, of course, gastronomical preferences. Cuisines are exemplars of human thinking and ideation process evolved over a number of civilizational constraints. While cuisines all over the world use many ingredients in common, they are distinguished by the patterns of usage of ingredients. Every cuisine can be seen as network whose nodes are the ingredients used in its recipes with links indicating which ingredients are used together. In this thesis, several cuisine networks are constructed incorporating semantic relevance between nodes using data mining and natural language processing techniques. These networks are compared in terms of their structural and statistical characteristics using techniques from graph theory and multivariate analysis. This includes calculating features such as degree distributions, clustering, diameters, assortativity, etc., as well as building correlation and co-occurrence matrices of ingredients. The features obtained from this analysis are used to organize the cuisines into a cluster hierarchy, thus clarifying the relationships between them. A set of algorithms based on machine learning and graph theoretic techniques is developed to build a culinary evaluation and recommendation system called Critic which can check the degree to which a given set of ingredients is consistent within a given cuisine for a culinary recipe. Such a system can serve as the core of a generic ideation system for generating new (recipe) ideas and for classifying given (recipes) ideas systematically into specific (cuisines) contexts.
Advisors/Committee Members: Minai, Ali.
Subjects: Computer Science
Keywords: cuisines; complex networks; semantic analysis; machine learning; ideation; information retreival
More Like This
23.
Wang, Mengxia.
Design of a Modified P300 Speller System Based on Prediction by Partial Matching Language Model.
Degree: MS, Engineering and Applied Science: Computer Science, 2012, University of Cincinnati
► In recent decades, the field of Brain-Computer Interface (BCI) technologies has been…
(more)
▼ In recent decades, the field of Brain-Computer Interface (BCI) technologies has been vigorously developed by research groups from all over the world. A BCI system can build a directly pathway between the brain and an external device so that patients with impaired motor activities can be assisted with this system to realize the communication with others. Among varieties of BCI systems, P300 speller is one that has been successfully developed with several advantages such as easy to carry on the experiment and the achievement of relatively better accuracy rate. In this thesis a system is designed to improve the current P300 speller performance based on a language prediction model. The method is implemented with MATLAB simulations in the evaluation of both system accuracy and speedup. The result of the conducted experiments shows the feasibility of our proposed system.
Advisors/Committee Members: Ralescu, Anca.
Subjects: Computer Science
Keywords: Brain-computer interface; P300 speller; Prediction by partial matching language model
More Like This
