Engineering topics
Lu, Hao
Publications and source records attributed to Lu, Hao.
Algorithms for Balanced Graph Colorings with Applications in Parallel Computing
Graph coloring is used to identify subsets of independent tasks in parallel scientific computing applications. Traditional coloring heuristics aim to reduce the number of colors used as that number also corresponds to the number of parallel steps in the application. However, if the color classes produced have a skew in their sizes, utilization of hardware resources becomes inefficient, especially for the smaller color classes. Equitable coloring is a theoretical formulation of coloring that guarantees a perfect balance among color classes, and its practical relaxation is referred to as balanced coloring. In this paper, we revisit the problem of balanced coloring in the context of parallel computing. The goal is to achieve a balanced coloring of an input graph without increasing the number of colors that an algorithm oblivious to balance would have used. We propose and study multiple heuristics that aim to achieve such a balanced coloring for two variants of coloring problem, distance-1 coloring and partial distance-2 coloring (defined on a bipartite graph). We present parallelization approaches for multi-core and manycore architectures, and cross-evaluate their effectiveness with respect to the quality of balance achieved and performance. Furthermore, we study the impact of the proposed balanced coloring heuristics on a concrete application - viz. parallel community detection, which is an example of an irregular application. In addition, we propose several extensions to our balancing schemes and evaluate their balancing efficacy and performance characteristics. The thorough treatment of balanced coloring presented in this paper from algorithms to application is expected to serve as a valuable resource to parallel application developers who seek to improve parallel performance of their applications using coloring.
Fast Uncovering of Graph Communities on a Chip: Toward Scalable Community Detection on Multicore and Manycore Platforms
Graph representations are pervasive in scientific and social computing. They serve as vital tools to model the interplay among different interacting entities. In this paper, we visit the problem of community detection, which is one of the most widely used graph operations toward scientific discovery. Community detection refers to the process of identifying tightly-knit subgroups of vertices in a large graph. These sub-groups (or communities) represent vertices that are tied together through common structure or function. Identification of communities could help in understanding the modular organization of complex networks. However, owing to large data sizes and high computational costs, performing community detection at scale has become increasingly challenging. Here, we present a detailed review and analysis of some of the leading computational methods and implementations developed for executing community detection on modern day multicore and manycore architectures. Our goals are to: a) define the problem of community detection and highlight its scientific significance; b) relate to challenges in parallelizing the operation on modern day architectures; c) provide a detailed report and taxonomical classification of the approaches that have been designed for various architectures; and d) finally, provide insights into the strengths and suitability of different architectures for community detection, and a preview into the future trends of the area. It is our hope that this detailed treatment of community detection on parallel architectures can serve as an exemplar study for extending the application of modern day multicore and manycore architectures to other complex graph applications.
Crystal structures of apo and inhibitor-bound TGFβR2 kinase domain: insights into TGFβR isoform selectivity
The cytokine TGF-β modulates a number of cellular activities and plays a critical role in development, hemostasis and physiology, as well as in diseases including cancer and fibrosis. TGF-β signals through two transmembrane serine/threonine kinase receptors: TGFβR1 and TGFβR2. Multiple structures of the TGFβR1 kinase domain are known, but the structure of TGFβR2 remains unreported. Wild-type TGFβR2 kinase domain was refractory to crystallization, leading to the design of two mutated constructs: firstly, a TGFβR1 chimeric protein with seven ATP-site residues mutated to their counterparts in TGFβR2, and secondly, a reduction of surface entropy through mutation of six charged residues on the surface of the TGFβR2 kinase domain to alanines. These yielded apo and inhibitor-bound crystals that diffracted to high resolution (<2 Å). Comparison of these structures with those of TGFβR1 reveal shared ligand contacts as well as differences in the ATP-binding sites, suggesting strategies for the design of pan and selective TGFβR inhibitors.
Understanding the bond-energy, hardness, and adhesive force from the phase diagram via the electron work function
Properties of metallic materials are intrinsically determined by their electron behavior. However, relevant theoretical treatment involving quantum mechanics is complicated and difficult to be applied in materials design. Electron work function (EWF) has been demonstrated to be a simple but fundamental parameter which well correlates properties of materials with their electron behavior and could thus be used to predict material properties from the aspect of electron activities in a relatively easy manner. In this article, we propose a method to extract the electron work functions of binary solid solutions or alloys from their phase diagrams and use this simple approach to predict their mechanical strength and surface properties, such as adhesion. Two alloys, Fe-Ni and Cu-Zn, are used as samples for the study. EWFs extracted from phase diagrams show same trends as experimentally observed ones, based on which hardness and surface adhesive force of the alloys are predicted. This new methodology provides an alternative approach to predict material properties based on the work function, which is extractable from the phase diagram. This work may also help maximize the power of phase diagram for materials design and development.
Dependence of the mechanical behavior of alloys on their electron work function—An alternative parameter for materials design
In this article, we demonstrate that the electron work function (EWF) as an intrinsic parameter can provide information or clues in a simple or straightforward way for material design, modification, and development. A higher work function of a material represents a more stable electronic state, which consequently generates a higher resistance to any attempt of changing the electronic state and other corresponding states, e.g., changes in structure or microstructure caused by mechanical and electrochemical actions. Using Cu-Ni alloy as an example, we demonstrate the correlation between the EWF and Young's modulus of the material as well as its hardness. The properties of a material can be modified using elements with appropriate work functions. This is also applicable for tailoring inter-phase boundaries or interfaces.