Search

Natural gradient method provides a powerful paradigm for training statistical models and offers several appealing theoretic benefits. It constructs the Fisher information matrix to correct the ordinary gradients, and thus, the cost may become prohibitively expensive in the large-scale setting. This paper proposes a quasi-natural gradient method to address this computational issue. It employs a rank-one model to capture the second-order information from the underlying statistical manifold and to iteratively update the Fisher approximations. A three-loop procedure is designed to implement the updating formulas. This procedure resembles the classical two-loop procedure in the limited-memory BFGS method but saves half of the memory usage while it can be made faster. The resulting method retains the convergence rate advantages of existing stochastic optimization methods and has inherent ability in handling nonconvexity. Numerical studies conducted on several machine learning tasks demonstrate the reduction in convergence time and the robustness in tackling nonconvexity relative to stochastic gradient descent and the online limited-memory BFGS method. [ABSTRACT FROM AUTHOR]

Previous studies including ours demonstrated that methyl-CpG-binding domain 2 (MBD2) acts as a reader to decipher DNA methylome-encoded information. We thus in the current study used Mbd22/2 mice as a model to dissect the impact of high-fat diet (HFD) on DNA methylome relevant to the pathoetiology of obesity. It was interestingly noted that mice deficient in Mbd2 were protected from HFD-induced obesity and insulin resistance. Mechanistic study revealed that HFD rendered epididymal adipose tissues to undergo a DNA methylation turnover as evidenced by the changes of methylation levels and patterns. Specifically, HFD was noted with higher potency to induce DNA hypomethylation in genes relevant to energy storage than that in genes associated with energy expenditure. As a result, arrays of genes were subjected to expression changes, which led to an altered homeostasis for energy storage and expenditure in favor of obesity development. Loss of Mbd2 resulted in impaired implementation of above DNA methylation changes associated with altered energy homeostasis, which then protected mice from HFD-induced obesity and insulin resistance. Those data would provide novel insight into the understanding of the pathoetiology underlying obesity with potential for developing effective therapies against obesity in clinical settings. [ABSTRACT FROM AUTHOR]

Let PG(k) be the number of proper k-colorings of a finite simple graph G. Tomescu's conjecture, which was recently solved by Fox, He, and Manners, states that PG(k) leq k!(k 1)n k for all connected graphs G on n vertices with chromatic number kgeq 4. In this paper, we study the same problem with the additional constraint that G is ell -connected. For 2-connected graphs G, we prove a tight bound PG(k) leq (k 1)!((k 1)n k+1 + (1)n k) and show that equality is only achieved if G is a k-clique with an ear attached. For ell geq 3, we prove an asymptotically tight upper bound PG(k) leq k!(k 1)n ell k+1 + O((k 2)n) and provide a matching lower bound construction. For the ranges kgeqll or ellgeq (k 2)(k 1) + 1 we further find the unique graph maximizing PG(k). We also consider generalizing ell -connected graphs to connected graphs with minimum degree delta. [ABSTRACT FROM AUTHOR]

On the basis that ground motions may arrive at a structure from any horizontal direction and that different directions of seismic incidence would result in different structural dynamic responses, this paper focuses on orienting the crucial seismic incidence of transmission tower-line systems based on the wavelet energy method. A typical transmission tower-line system is chosen as the case study, and two finite element (FE) models are established in ABAQUS, with and without consideration of the interaction between the transmission towers and the transmission lines. The mode combination frequency is defined by considering the influence of the higher-order modes of the structure. Subsequently, wavelet transformation is performed to obtain the total effective energy input and the effective energy input rate corresponding to the mode combination frequency to further judge the critical angle of seismic incidence by comparing these two performance indexes under different seismic incidence angles. To validate this approach, finite element history analysis (FEHA) is imposed on both FE models to generate comparative data, and good agreement is found. The results demonstrate that the wavelet energy method can forecast the critical angle of seismic incidence of a transmission tower-line system with adequate accuracy, avoiding time-consuming and cumbersome computer analysis. The proposed approach can be used in future seismic design of transmission tower-line systems. [ABSTRACT FROM AUTHOR]

Background: This study is to explore the key genes and signaling transduction pathways related to the survival time of glioblastoma multiforme (GBM) patients. Results: Our results not only showed that mutually explored GBM survival time related genes and signaling transduction pathways are closely related to the GBM, but also demonstrated that our innovated constrained optimization algorithm (CoxSisLasso strategy) are better than the classical methods (CoxLasso and CoxSis strategy). Conclusion: We analyzed why the CoxSisLasso strategy can outperform the existing classical methods and discuss how to extend this research in the distant future. [ABSTRACT FROM AUTHOR]

Background: Despite decades of extensive studies, the morbidity and mortality for acute lung injury/acute respiratory distress syndrome (ALI/ARDS) remained high. Particularly, biomarkers essential for its early diagnosis and prognosis are lacking. Methods: Recent studies suggest that alveolar macrophages (AMs) at the exudative phase of ALI/ARDS initiate, amplify and perpetuate inflammatory responses, while they resolve inflammation in the recovery phase to prevent further tissue injury and perpetuated inflammation in the lung. Therefore, proteins relevant to this functional switch could be valuable biomarkers for ALI/ARDS diagnosis and prognosis. We thus conducted comparative analysis of the AM proteome to assess its dynamic proteomic changes during ALI/ARDS progression and recovery. Results: 135 proteins were characterized to be differentially expressed between AMs at the exudative and recovery phase. MALDI-TOF-MS and peptide mass fingerprint (PMF) analysis characterized 27 informative proteins, in which 17 proteins were found with a marked increase at the recovery phase, while the rest of 10 proteins were manifested by the significantly higher levels of expression at the exudative phase. Conclusions: Given the role of above identified proteins played in the regulation of inflammatory responses, cell skeleton organization, oxidative stress, apoptosis and metabolism, they have the potential to serve as biomarkers for early diagnosis and prognosis in the setting of patients with ALI/ARDS. [ABSTRACT FROM AUTHOR]