A comparative study on large scale kernelized support vector machines

Kernelized support vector machines (SVMs) belong to the most widely used classification methods. However, in contrast to linear SVMs, the computation time required to train such a machine becomes a bottleneck when facing large data sets. In order to mitigate this shortcoming of kernel SVMs, many approximate training algorithms were developed. While most of these methods claim to be much faster than the state-of-the-art solver LIBSVM, a thorough comparative study is missing. We aim to fill this gap. We choose several well-known approximate SVM solvers and compare their performance on a number of large benchmark data sets. Our focus is to analyze the trade-off between prediction error and runtime for different learning and accuracy parameter settings. This includes simple subsampling of the data, the poor-man’s approach to handling large scale problems. We employ model-based multi-objective optimization, which allows us to tune the parameters of learning machine and solver over the full range of accuracy/runtime trade-offs. We analyze (differences between) solvers by studying and comparing the Pareto fronts formed by the two objectives classification error and training time. Unsurprisingly, given more runtime most solvers are able to find more accurate solutions, i.e., achieve a higher prediction accuracy. It turns out that LIBSVM with subsampling of the data is a strong baseline. Some solvers systematically outperform others, which allows us to give concrete recommendations of when to use which solver.     

Quasicrystalline electronic states of a one-dimensionally modulated Ag film

Ag films on GaAs(110) exhibit a one-dimensional quasiperiodic modulation, resulting in a Fibonacci sequence of parallel stripes with two different widths. Valence level photoemission shows that the Ag electronic states acquire a unique character along the quasiperiodic direction, distinctively manifested by a hierarchy of energy level replicas and avoided crossings at characteristic intervals in reciprocal space. These observations demonstrate the strong influence of the one-dimensional quasiperiodic potential on the Ag film states.     

Turnaround Students in High School Mathematics: Constructing Identities of Competence Through Mathematical Worlds

This study investigates prospective secondary teachers’ cognitive difficulties and mathematical ideas involved in making connections among representations. We implemented a three-week teaching unit to help prospective secondary mathematics teachers develop understanding of big ideas that are critical to formulating connections among representations, in the context of conic curves. Qualitative analysis of data showed that most undergraduate mathematics majors and minors in this study struggled with variation, the Cartesian Connection, and other affiliated ideas such as graph as a locus of points. Furthermore, they were unable to identify basic metric relations encoded in algebraic expressions such as the distance between points, which further compounded their difficulties in making connections among representations. We argue that mathematics teacher education needs more focus on these ideas so that their graduates can successfully teach these big ideas in their future instruction.     

Non-interferometric transient quantitative phase microscopy for ultrafast engineering

Lead-free piezoelectric ceramics Bi_0.5(Na_1-x-yK_xAg_y)_0.5TiO₃ [BNKAT(x/y)] have been synthesized by the mixed oxide method. The effects of the amount of K⁺ and Ag⁺ on the electrical properties were examined. X-ray diffraction patterns indicate that K⁺ and Ag⁺ ions partially substitute for the Na⁺ ions in Bi_0.5Na_0.5TiO₃ and form a solid solution during sintering. At room temperature, the ceramics exhibit good performances with piezoelectric constant d₃₃=189 pC/N, electromechanical coupling factor k_p=35.0%, remanent polarization P_r=39.5 μC/cm², and coercive field E_c=3.3 kV/mm, respectively. The curves of the dielectric constant ε_r and loss tangent tan δ versus temperature show that the transition temperature from ferroelectric to anti-ferroelectric phase decreases with increasing the K⁺ content for the compositions researched. The dependencies of k_p and polarization versus electric (P–E) hysteresis loops on temperature reveal that the depolarization temperature T_d of BNKAT(0.15/0.015) ceramics, which have good piezoelectric properties (d₃₃=134 pC/N, k_p=32.5%) and strong ferroelectricity (P_r=39.5 μC/cm², E_c=4.1 kV/mm) at room temperature, is above 160 °C. PACS 77.22.-d; 77.65.Bn; 77.80.Bh; 77.80.Dj; 77.84.Dy     

Evaluation of Co and Zn competitive sorption by zeolitic material synthesized from fly ash using 60Co and 65Zn as radioindicators

Journal of Radioanalytical and Nuclear Chemistry - Zeolitic material prepared by hydrothermal treatment of brown coal fly ash was used for sorption separation of Co2+ and Zn2+ ions from single and...     

Measuring surface forces to explore surface chemistry: Mica, sapphire and silica

In this paper measurements of the forces acting between two solid surfaces separated by a thin liquid film are discussed. By investigating these forces in a range of different liquids and solutions, it is possible to acquire an understanding of the surface properties of the solid material. The surface of mica has been studied extensively in this way, and the results obtained are reviewed to illustrate how the surface force measurements can give surface chemical information. Recent measurements on two other materials, sapphire and silica, which are of greater practical interest are also discussed.     

Isomorphic edge disjoint subgraphs of hypergraphs

We show that any k‐uniform hypergraph with n edges contains two isomorphic edge disjoint subgraphs of size for k = 4, 5 and 6. This is best possible up to a logarithmic factor due to an upper bound construction of Erd?s, Pach, and Pyber who show there exist k‐uniform hypergraphs with n edges and with no two edge disjoint isomorphic subgraphs with size larger than . Furthermore, our result extends results Erd?s, Pach and Pyber who also established the lower bound for k = 2 (eg. for graphs), and of Gould and Rödl who established the result for k = 3. © 2016 Wiley Periodicals, Inc. Random Struct. Alg., 48, 767–793, 2016     

Experimental and Computational Analysis of the Solvent‐Dependent O2/Li+‐O2− Redox Couple: Standard Potentials,Coupling Strength,and Implications for Lithium–Oxygen Batteries

Understanding and controlling the kinetics of O₂ reduction in the presence of Li⁺‐containing aprotic solvents, to either Li⁺‐O₂⁻ by one‐electron reduction or Li₂O₂ by two‐electron reduction, is instrumental to enhance the discharge voltage and capacity of aprotic Li‐O₂ batteries. Standard potentials of O₂/Li⁺‐O₂⁻ and O₂/O₂⁻ were experimentally measured and computed using a mixed cluster‐continuum model of ion solvation. Increasing combined solvation of Li⁺ and O₂⁻ was found to lower the coupling of Li⁺‐O₂⁻ and the difference between O₂/Li⁺‐O₂⁻ and O₂/O₂⁻ potentials. The solvation energy of Li⁺ trended with donor number (DN), and varied greater than that of O₂⁻ ions, which correlated with acceptor number (AN), explaining a previously reported correlation between Li⁺‐O₂⁻ solubility and DN. These results highlight the importance of the interplay between ion–solvent and ion–ion interactions for manipulating the energetics of intermediate species produced in aprotic metal–oxygen batteries.