Theses and Dissertations
Permanent URI for this collection
Browse
Browsing Theses and Dissertations by Author "Anderson, Chuck, committee member"
Now showing 1 - 3 of 3
Results Per Page
Sort Options
Item Open Access Laplacian Eigenmaps for time series analysis(Colorado State University. Libraries, 2020) Rosse, Patrick J., author; Kirby, Michael, advisor; Peterson, Chris, committee member; Adams, Henry, committee member; Anderson, Chuck, committee memberWith "Big Data" becoming more available in our day-to-day lives, it becomes necessary to make meaning of it. We seek to understand the structure of high-dimensional data that we are unable to easily plot. What shape is it? What points are "related" to each other? The primary goal is to simplify our understanding of the data both numerically and visually. First introduced by M. Belkin, and P. Niyogi in 2002, Laplacian Eigenmaps (LE) is a non-linear dimensional reduction tool that relies on the basic assumption that the raw data lies in a low-dimensional manifold in a high-dimensional space. Once constructed, the graph Laplacian is used to compute a low-dimensional representation of the data set that optimally preserves local neighborhood information. In this thesis, we present a detailed analysis of the method, the optimization problem it solves, and we put it to work on various time series data sets. We show that we are able to extract neighborhood features from a collection of time series, which allows us to cluster specific time series based on noticeable signatures within the raw data.Item Open Access Mean variants on matrix manifolds(Colorado State University. Libraries, 2012) Marks, Justin D., author; Peterson, Chris, advisor; Kirby, Michael, advisor; Bates, Dan, committee member; Anderson, Chuck, committee memberThe geometrically elegant Stiefel and Grassmann manifolds have become organizational tools for data applications, such as illumination spaces for faces in digital photography. Modern data analysis involves increasingly large-scale data sets, both in terms of number of samples and number of features per sample. In circumstances such as when large-scale data has been mapped to a Stiefel or Grassmann manifold, the computation of mean representatives for clusters of points on these manifolds is a valuable tool. We derive three algorithms for determining mean representatives for a cluster of points on the Stiefel manifold and the Grassmann manifold. Two algorithms, the normal mean and the projection mean, follow the theme of the Karcher mean, relying upon inversely related maps that operate between the manifold and the tangent bundle. These maps are informed by the geometric definition of the tangent bundle and the normal bundle. From the cluster of points, each algorithm exploits these maps in a predictor/corrector loop until converging, with prescribed tolerance, to a fixed point. The fixed point acts as the normal mean representative, or projection mean representative, respectively, of the cluster. This method shares its principal structural characteristics with the Karcher mean, but utilizes a distinct pair of inversely related maps. The third algorithm, called the flag mean, operates in a context comparable to a generalized Grassmannian. It produces a mean subspace of arbitrary dimension. We provide applications and discuss generalizations of these means to other manifolds.Item Open Access Storing cycles in Hopfield-type neural networks(Colorado State University. Libraries, 2014) Zhang, Chuan, author; Dangelmayr, Gerhard, advisor; Oprea, Iuliana, advisor; Shipman, Patrick, committee member; Anderson, Chuck, committee memberThe storage of pattern sequences is one of the most important tasks in both biological and artificial intelligence systems. Clarifying the underlying mathematical principles for both the storage and retrieval of pattern sequences in neural networks is fundamental for understanding the generation of rhythmic movements in animal nervous systems, as well as for designing electrical circuits to produce and control rhythmic output. In this dissertation, we investigate algebraic structures of binary cyclic patterns (or for short cycles) and study relations between these structures and the topology and dynamics of Hopfield-type networks with and without delay constructed from cyclic patterns using the pseudoinverse learning rule. A cycle defined by a binary matrix Σ is called admissible, if a connectivity matrix J satisfying the cycle's transition conditions exists. We show that Σ is admissible, if and only if its discrete Fourier transform contains exactly r = rank(Σ) nonzero columns. Based on the decomposition of the rows of Σ into disjoint subsets corresponding to loops, where a loop is defined by the set of all cyclic permutations of a row, cycles are classified as simple cycles, and separable or inseparable composite cycles. Simple cycles contain rows from one loop only, and the network topology is a feed-forward chain with feedback to one neuron if the loop-vectors in Σ are cyclic permutations of each other. For special cases this topology simplifies to a ring with only one feedback. Composite cycles contain rows from at least two disjoint loops, and the neurons corresponding to the loop-vectors in Σ from the same loop are identified with a cluster. Networks constructed from separable composite cycles decompose into completely isolated clusters. For inseparable composite cycles at least two clusters are connected, and the cluster-connectivity is related to the intersections of the spaces spanned by the loop-vectors of the clusters. The remainder of this thesis deals with the dynamics of Hopfield-type networks with connectivities constructed from admissible cycles. In this approach, the connectivity is composed of two contributions, C0J0 and C1J, where the matrix J0 serves to store cycle's patterns as fixed points and the matrix J induces the transitions between the cycle's patterns. Delayed couplings are associated with the transition matrix J. An admissible cycle is called strongly retrievable if for appropriate initial data the network dynamics undergoes a persistent oscillation in accordance with cycle's transition conditions. An admissible cycle is called weakly retrievable if for any M there exists a sufficiently large delay time τ such that at least M consecutive patterns are retrieved. When the Hamming distance between successive cycle-patterns is greater than one, the sign-changes in the network dynamics occur asynchronously, leading to the occurrence of intermediate patterns that are not contained in the cycle-matrix. We call the time-intervals with these intermediate patterns misalignment intervals and introduce a novel method to analyze the lengths of these intervals, which is referred to as Misalignment Length Analysis (MLA). Using this method, intermediate patterns are determined and for a special class of cycles a recurrence relation for successive misalignment intervals is established. In addition, a class of cycles, related to properties of the intermediate patterns, is identified which can be proved to be weakly retrievable in the case C0 = 0 and for sufficiently large values of the gain scaling parameter, λ, of the sigmoid coupling function. More generally, we also prove that for a given J constructed from a preselected cycle in that class, all other cycles satisfying the transition conditions associated with J are weakly retrievable as well. These results provide an analytic explanation for the long-lasting transient oscillations observed recently in simulations of cooperative Hopfield-type networks with delays. For general values of C0, C1, λ, we perform a linear stability analysis and give a complete description of all possible bifurcations of the trivial solution for networks constructed from admissible cycles. Numerically we illustrate that, depending on the structural features of a cycle, admissible cycles are stored and retrieved either as attracting limit cycles or as long-lasting transient oscillations. Moreover, if the cycle is revealed as attracting limit cycle, this limit cycle is created in a Hopf bifurcation from the trivial solution, and the transition from fixed point attractors to the attracting limit cycle is established through multiple saddle-nodes on limit cycle bifurcations. Lastly, simulations showing successfully retrieved cycles in continuous-time Hopfield-type networks and in networks of spiking neurons exhibiting up-down states are presented, which strongly suggests that the results of the study presented in this dissertation can be extended to more complicated networks.