THE KAUST Repository is an initiative of the University Library to expand the impact of conference papers, technical reports, peer-reviewed articles, preprints, theses, images, data sets, and other research-related works of King Abdullah University of Science and Technology (KAUST).
Files in the repository are accessible through popular web search engines and are given persistent web addresses so links will not become broken over time.
KAUST researchers: To submit your research to the repository, click on Submit an Item, log in with your KAUST user name and password, and deposit the item in the appropriate collection.
If you have any questions, please contact firstname.lastname@example.org.
Urban Image Analysis with Convolutional Sparse Coding(2018-09-18)Urban image analysis is one of the most important problems lying at the intersection of computer graphics and computer vision research. In addition, Convolutional Sparse Coding (CSC) is a well-established image representation model especially suited for image restoration tasks. This dissertation handles urban image analysis using an asset extraction framework, studies CSC for the reconstruction of both urban and general images using supervised data, and proposes a better computational approach to CSC. Our asset extraction framework uses object proposals which are currently used for increasing the computational efficiency of object detection. In this dissertation, we propose a novel adaptive pipeline for interleaving object proposals with object classification and use it as a formulation for asset detection. We first preprocess the images using a novel and efficient rectification technique. We then employ a particle filter approach to keep track of three priors, which guide proposed samples and get updated using classifier output. Tests performed on over 1000 urban images demonstrate that our rectification method is faster than existing methods without loss in quality, and that our interleaved proposal method outperforms current state-of-the-art. We further demonstrate that other methods can be improved by incorporating our interleaved proposals. We also extend the applicability of the CSC model by proposing a supervised approach to the problem, which aims at learning discriminative dictionaries instead of purely reconstructive ones. We incorporate a supervised regularization term into the traditional unsupervised CSC objective to encourage the final dictionary elements to be discriminative. Experimental results show that using supervised convolutional learning results in two key advantages. First, we learn more semantically relevant filters in the dictionary and second, we achieve improved image reconstruction on unseen data. We finally present two computational contributions to the state of the art in CSC. First, we significantly speed up the computation by proposing a new optimization framework that tackles the problem in the dual domain. Second, we extend the original formulation to higher dimensions in order to process a wider range of inputs, such as RGB images and videos. Our results show up to 20 times speedup compared to current state-of-the-art CSC solvers.
Three-dimensional Modeling and Simulation of a Tuning Fork(2018-09-16)The mathematical characterization of the sound of a musical instrument still follows Schumann’s laws . According to this theory, the resonances of the instrument body, “the formants”, filter the oscillations of the sound generator (e.g., strings) and produce the characteristic “timbre” of an instrument. This is a strong simplification of the actual situation. It applies to a point source and does not distinguish between a loudspeaker and a three-dimensional instrument. In this work we investigate Finite-Element-based numerical simulations of eigenfrequencies and eigenmodes of a tuning fork in order to capture the oscillation behavior of its eigenfrequencies. We model the tuning fork as an elastic solid body and solve an eigenvalue equation derived from a system of coupled equations from linear elasticity theory on an unstructured three-dimensional grid. The eigenvalue problem is solved using the preconditioned inverse iteration (PINVIT) method with an efficient geometric multigrid (GMG) preconditioner. The latter allows us to resolve the tuning fork with a high resolution grid, which is required to capture fine modes of the simulated eigenfrequencies. To verify our results, we compare them with measurement data obtained from an experimental modal analyses of a real reference tuning fork. It turns out that our model is sufficient to capture the first eight eigenmodes of a reference tuning fork, whose identification and reproduction by simulation is novel to the knowledge of the author.
Behavior of numerical error in pore-scale lattice Boltzmann simulations with simple bounce-back rule: Analysis and highly accurate extrapolation(AIP Publishing, 2018-09-14)We perform the viscosity-independent Stokes flow simulations in regular sphere packings using the two-relaxation-times (TRT) lattice Boltzmann method (LBM) with the simple bounce-back (BB) rule. Our special discretization procedure reduces the scatter in integral quantities, such as drag force, and quantifies the solution convergence error. We assume transition to linear (−1) convergence rate for different sets of TRT parameters and use this assumption to provide a simple extrapolation scheme. After establishing the accurate reference values of drag for a wide range of porosities, 0.26–0.78, we show a ten-fold decrease in the drag error using the suggested extrapolations. This error decrease allows the simple LBM/BB scheme to reach an accuracy of the high-order interpolated boundary schemes. The suggested extrapolation approach is straightforward to apply in porous media, whose pore space can be discretized at several resolutions.
Vocal practice regulates singing activity–dependent genes underlying age-independent vocal learning in songbirds(Public Library of Science (PLoS), 2018-09-12)The development of highly complex vocal skill, like human language and bird songs, is underlain by learning. Vocal learning, even when occurring in adulthood, is thought to largely depend on a sensitive/critical period during postnatal development, and learned vocal patterns emerge gradually as the long-term consequence of vocal practice during this critical period. In this scenario, it is presumed that the effect of vocal practice is thus mainly limited by the intrinsic timing of age-dependent maturation factors that close the critical period and reduce neural plasticity. However, an alternative, as-yet untested hypothesis is that vocal practice itself, independently of age, regulates vocal learning plasticity. Here, we explicitly discriminate between the influences of age and vocal practice using a songbird model system. We prevented zebra finches from singing during the critical period of sensorimotor learning by reversible postural manipulation. This enabled to us to separate lifelong vocal experience from the effects of age. The singing-prevented birds produced juvenile-like immature song and retained sufficient ability to acquire a tutored song even at adulthood when allowed to sing freely. Genome-wide gene expression network analysis revealed that this adult vocal plasticity was accompanied by an intense induction of singing activity-dependent genes, similar to that observed in juvenile birds, rather than of age-dependent genes. The transcriptional changes of activity-dependent genes occurred in the vocal motor robust nucleus of the arcopallium (RA) projection neurons that play a critical role in the production of song phonology. These gene expression changes were accompanied by neuroanatomical changes: dendritic spine pruning in RA projection neurons. These results show that self-motivated practice itself changes the expression dynamics of activity-dependent genes associated with vocal learning plasticity and that this process is not tightly linked to age-dependent maturational factors.