Hydrological Response of East China to the Variation of East Asian Summer Monsoon

Copyright © 2016 Fuxing Li et al.The sensitivity of hydrologic variables in East China, that is, runoff, precipitation, evapotranspiration, and soil moisture to the fluctuation of East Asian summer monsoon (EASM), is evaluated by the Mann-Kendall correlation analysis on a spatial resolution of 1/4° in the period of 1952-2012. The results indicate remarkable spatial disparities in the correlation between the hydrologic variables and EASM. The regions in East China susceptible to hydrological change due to EASM fluctuation are identified. When the standardized anomaly of intensity index of EASM (EASMI) is above 1.00, the runoff of Haihe basin has increased by 49% on average, especially in the suburb of Beijing and Hebei province where the runoff has increased up to 105%. In contrast, the runoff in the basins of Haihe and Yellow River has decreased by about 27% and 17%, respectively, when the standardized anomaly of EASMI is below -1.00, which has brought severe drought to the areas since mid-1970s. The study can be beneficial for national or watershed agencies developing adaptive water management strategies in the face of global climate change.

Learning from Geometry

Subspaces and manifolds are two powerful models for high dimensional signals. Subspaces model linear correlation and are a good fit to signals generated by physical systems, such as frontal images of human faces and multiple sources impinging at an antenna array. Manifolds model sources that are not linearly correlated, but where signals are determined by a small number of parameters. Examples are images of human faces under different poses or expressions, and handwritten digits with varying styles. However, there will always be some degree of model mismatch between the subspace or manifold model and the true statistics of the source. This dissertation exploits subspace and manifold models as prior information in various signal processing and machine learning tasks.

A near-low-rank Gaussian mixture model measures proximity to a union of linear or affine subspaces. This simple model can effectively capture the signal distribution when each class is near a subspace. This dissertation studies how the pairwise geometry between these subspaces affects classification performance. When model mismatch is vanishingly small, the probability of misclassification is determined by the product of the sines of the principal angles between subspaces. When the model mismatch is more significant, the probability of misclassification is determined by the sum of the squares of the sines of the principal angles. Reliability of classification is derived in terms of the distribution of signal energy across principal vectors. Larger principal angles lead to smaller classification error, motivating a linear transform that optimizes principal angles. This linear transformation, termed TRAIT, also preserves some specific features in each class, being complementary to a recently developed Low Rank Transform (LRT). Moreover, when the model mismatch is more significant, TRAIT shows superior performance compared to LRT.

The manifold model enforces a constraint on the freedom of data variation. Learning features that are robust to data variation is very important, especially when the size of the training set is small. A learning machine with large numbers of parameters, e.g., deep neural network, can well describe a very complicated data distribution. However, it is also more likely to be sensitive to small perturbations of the data, and to suffer from suffer from degraded performance when generalizing to unseen (test) data.

From the perspective of complexity of function classes, such a learning machine has a huge capacity (complexity), which tends to overfit. The manifold model provides us with a way of regularizing the learning machine, so as to reduce the generalization error, therefore mitigate overfiting. Two different overfiting-preventing approaches are proposed, one from the perspective of data variation, the other from capacity/complexity control. In the first approach, the learning machine is encouraged to make decisions that vary smoothly for data points in local neighborhoods on the manifold. In the second approach, a graph adjacency matrix is derived for the manifold, and the learned features are encouraged to be aligned with the principal components of this adjacency matrix. Experimental results on benchmark datasets are demonstrated, showing an obvious advantage of the proposed approaches when the training set is small.

Stochastic optimization makes it possible to track a slowly varying subspace underlying streaming data. By approximating local neighborhoods using affine subspaces, a slowly varying manifold can be efficiently tracked as well, even with corrupted and noisy data. The more the local neighborhoods, the better the approximation, but the higher the computational complexity. A multiscale approximation scheme is proposed, where the local approximating subspaces are organized in a tree structure. Splitting and merging of the tree nodes then allows efficient control of the number of neighbourhoods. Deviation (of each datum) from the learned model is estimated, yielding a series of statistics for anomaly detection. This framework extends the classical {\em changepoint detection} technique, which only works for one dimensional signals. Simulations and experiments highlight the robustness and efficacy of the proposed approach in detecting an abrupt change in an otherwise slowly varying low-dimensional manifold.

Genetic Analysis of Mitotic Recombination in Saccharomyces cerevisiae

Mitotic genome instability can occur during the repair of double-strand breaks (DSBs) in DNA, which arise from endogenous and exogenous sources. Studying the mechanisms of DNA repair in the budding yeast, Saccharomyces cerevisiae has shown that Homologous Recombination (HR) is a vital repair mechanism for DSBs. HR can result in a crossover event, in which the broken molecule reciprocally exchanges information with a homologous repair template. The current model of double-strand break repair (DSBR) also allows for a tract of information to non-reciprocally transfer from the template molecule to the broken molecule. These “gene conversion” events can vary in size and can occur in conjunction with a crossover event or in isolation. The frequency and size of gene conversions in isolation and gene conversions associated with crossing over has been a source of debate due to the variation in systems used to detect gene conversions and the context in which the gene conversions are measured.

In Chapter 2, I use an unbiased system that measures the frequency and size of gene conversion events, as well as the association of gene conversion events with crossing over between homologs in diploid yeast. We show mitotic gene conversions occur at a rate of 1.3x10-6 per cell division, are either large (median 54.0kb) or small (median 6.4kb), and are associated with crossing over 43% of the time.

DSBs can arise from endogenous cellular processes such as replication and transcription. Two important RNA/DNA hybrids are involved in replication and transcription: R-loops, which form when an RNA transcript base pairs with the DNA template and displaces the non-template DNA strand, and ribonucleotides embedded into DNA (rNMPs), which arise when replicative polymerase errors insert ribonucleotide instead of deoxyribonucleotide triphosphates. RNaseH1 (encoded by RNH1) and RNaseH2 (whose catalytic subunit is encoded by RNH201) both recognize and degrade the RNA in within R-loops while RNaseH2 alone recognizes, nicks, and initiates removal of rNMPs embedded into DNA. Due to their redundant abilities to act on RNA:DNA hybrids, aberrant removal of rNMPs from DNA has been thought to lead to genome instability in an rnh201Δ background.

In Chapter 3, I characterize (1) non-selective genome-wide homologous recombination events and (2) crossing over on chromosome IV in mutants defective in RNaseH1, RNaseH2, or RNaseH1 and RNaseH2. Using a mutant DNA polymerase that incorporates 4-fold fewer rNMPs than wild type, I demonstrate that the primary recombinogenic lesion in the RNaseH2-defective genome is not rNMPs, but rather R-loops. This work suggests different in-vivo roles for RNaseH1 and RNaseH2 in resolving R-loops in yeast and is consistent with R-loops, not rNMPs, being the the likely source of pathology in Aicardi-Goutières Syndrome patients defective in RNaseH2.

Whole-cell-analysis of live cardiomyocytes using wide-field interferometric phase microscopy.

We apply wide-field interferometric microscopy techniques to acquire quantitative phase profiles of ventricular cardiomyocytes in vitro during their rapid contraction with high temporal and spatial resolution. The whole-cell phase profiles are analyzed to yield valuable quantitative parameters characterizing the cell dynamics, without the need to decouple thickness from refractive index differences. Our experimental results verify that these new parameters can be used with wide field interferometric microscopy to discriminate the modulation of cardiomyocyte contraction dynamics due to temperature variation. To demonstrate the necessity of the proposed numerical analysis for cardiomyocytes, we present confocal dual-fluorescence-channel microscopy results which show that the rapid motion of the cell organelles during contraction preclude assuming a homogenous refractive index over the entire cell contents, or using multiple-exposure or scanning microscopy.

Hyperspectral Image Classification with Nonlinear Methods

This thesis introduces two related lines of study on classification of hyperspectral images with nonlinear methods. First, it describes a quantitative and systematic evaluation, by the author, of each major component in a pipeline for classifying hyperspectral images (HSI) developed earlier in a joint collaboration [23]. The pipeline, with novel use of nonlinear classification methods, has reached beyond the state of the art in classification accuracy on commonly used benchmarking HSI data [6], [13]. More importantly, it provides a clutter map, with respect to a predetermined set of classes, toward the real application situations where the image pixels not necessarily fall into a predetermined set of classes to be identified, detected or classified with.

The particular components evaluated are a) band selection with band-wise entropy spread, b) feature transformation with spatial filters and spectral expansion with derivatives c) graph spectral transformation via locally linear embedding for dimension reduction, and d) statistical ensemble for clutter detection. The quantitative evaluation of the pipeline verifies that these components are indispensable to high-accuracy classification.

Secondly, the work extends the HSI classification pipeline with a single HSI data cube to multiple HSI data cubes. Each cube, with feature variation, is to be classified of multiple classes. The main challenge is deriving the cube-wise classification from pixel-wise classification. The thesis presents the initial attempt to circumvent it, and discuss the potential for further improvement.

Using Image Processing Methods to Improve the Detection of Buried Explosive Threats in GPR Data

Current state of the art techniques for landmine detection in ground penetrating radar (GPR) utilize statistical methods to identify characteristics of a landmine response. This research makes use of 2-D slices of data in which subsurface landmine responses have hyperbolic shapes. Various methods from the field of visual image processing are adapted to the 2-D GPR data, producing superior landmine detection results. This research goes on to develop a physics-based GPR augmentation method motivated by current advances in visual object detection. This GPR specific augmentation is used to mitigate issues caused by insufficient training sets. This work shows that augmentation improves detection performance under training conditions that are normally very difficult. Finally, this work introduces the use of convolutional neural networks as a method to learn feature extraction parameters. These learned convolutional features outperform hand-designed features in GPR detection tasks. This work presents a number of methods, both borrowed from and motivated by the substantial work in visual image processing. The methods developed and presented in this work show an improvement in overall detection performance and introduce a method to improve the robustness of statistical classification.