Random Mutation Hill Climbing Research Articles

EPAW-1.0 code for evolutionary optimization of PAW datasets especially for high-pressure applications

We present a bio-inspired stochastic optimization strategy that optimizes projector augmented wave (PAW) datasets, for a user-specified pressure range, to realize the highest possible accuracy in high-throughput density functional theory calculations within the framework of PAW method. We named the optimizer “Evolutionary Generator of projector augmented wave datasets” (EPAW-1.0). The self-learning evolutionary algorithms in EPAW-1.0 adaptively tune some of the PAW parameters (such as different radii, and reference energies) to generate evolutionary optimized PAW (EPAW) datasets. In the course of designing EPAW dataset with a specific pseudo partial waves and projectors generation scheme, the code keeps the user-specified electronic configuration unaltered and the augmentation radius (rc) on the verge of the user allowed maximum without resulting in sphere overlap. The EPAW-1.0 algorithm homes on to a soft, transferable and unified EPAW dataset using various measures including the equation of state (EoS) of standard elemental materials within a user-specified pressure range that allows probing ∼50% volume compression with respect to the equilibrium atomic volume (corresponding to the energy minimum). The measures used by the EPAW algorithm also can be used to balance the efficiency and accuracy of the dataset.Program Title: EPAW-1.0Program Files doi:http://dx.doi.org/10.17632/ms52ym7vcn.1Licensing provisions: GNU General Public License 3 (GPLv3)Programming languages: Fortran90, Fortran77, Python3.0, bash shell, gnuplot5.0.External routines/libraries: ATOMPAW, Quantum ESPRESSO, related linear algebra package.Nature of problem: EPAW-1.0 is a hybrid recipe [2] that documents an interdisciplinary research integrating evolutionary computing with density functional theory (DFT). It offers a partially automated and consistent route to generate evolutionary optimized PAW (EPAW) datasets that show uniform performance for simulations up to a predefined high pressure. In particular, EPAW-1.0 makes use of evolutionary algorithms [3–7], PAW dataset generator (ATOMPAW [8], for example) and electronic-structure calculation code (Quantum Espresso [9], for example) to generate efficient and transferable EPAW datasets. EPAW datasets provide optimal accuracy in solid-state ab initio calculations within the favorable PAW computational framework — very close to the precision of targeted all-electron full potential linearized Augmented-plane-wave (AE-FLAPW) approach. We set the EoS from WIEN2k [1] calculations as our target, but any reliable EoS can be in use. The better the target EoS, the better is the performance of the EPAW datasets. Similarly, electronic-structure calculation code and PAW dataset generator code can also be replaced.Solution Method: The implementation has two parts: a. Generating a diverse random initial population for EPAW-1.0 to start with; b. Generating EPAW dataset using EPAW-1.0.A hybrid method, combining an in-house evolutionary strategy named Completely Adaptive Random Mutation Hill Climbing [6] (CARMHC) with ATOMPAW [8] program, quickly optimizes a set (population) of npop solutions (PAW dataset descriptors) from which the iterative process of EPAW-1.0 starts. The optimization condition involves superimposing the logarithmic derivative curves of the all-electron and pseudo radial wave functions, simultaneously satisfying necessary constraints on logarithmic derivatives, partial waves, pseudo partial waves and projector basis sets [2,8,10–12]. To start with, the method generates a diverse initial population of npop solution vector (npop= cardinality of the GA’s population in EPAW-1.0) randomly in the neighborhood of the initial educated guess based on problem-specific knowledge from the ATOMPAW program and iteratively tunes the population to minimize the area under the logarithmic derivative curves.Once npop number of solution strings have been optimized, npop individuals are feed into the EPAW-1.0’s initial population. Equilibrium total energies are evaluated using Quantum Espresso distribution [9] and fitted to a finite strain expansion. Pressures are calculated using a 4-parameter Birch–Murnaghan fit [13,14]. EPAW-1.0 employs genetic algorithms (GAs) [15–19] as the evolutionary procedure and iteratively tunes the free parameters of the PAW dataset generator to minimize the difference in equation of states (EoS) with respect to the given target EoS for the specified pressure range.Additional comments: EPAW-1.0 is an evolutionary procedure blended with external density functional theory (DFT) formalism (as implemented in ATOMPAW and Quantum Espresso, for example). The ATOMPAW code generates dataset by a self-consistent all-electron atomic structure calculation within the framework of DFT. The projector and basis functions are derived from the eigenstates of the all-electron atomic Hamiltonian. They are determined by iteratively solving radial differential equations. Equilibrium total energies at some equidistant volume points for a specific elemental crystal are evaluated using Quantum Espresso distribution [9]. The EPAW-1.0 program conserves different constraints [2] on logarithmic derivatives and basis sets to avoid numerical instability, ghost states, and to promote an excellent transferability. More details about the constraints can be found in the methodology part of reference 1. It also provides a goodness measure of the generated dataset concerning the targeted results.

RMHC-MR: Instance selection by random mutation hill climbing algorithm with MapReduce in big data

Instance selection is used to reduce the size of training set by removing redundant, erroneous and noisy instances and is an important pre-processing step in KDD (knowledge discovery in databases). Recently, to process very large data set, several methods divide the training set into disjoint subsets and apply instance selection algorithms to each subset independently. In this paper, we analyze the limitation of these methods and give our viewpoint about how to “divide and conquer” in instance selection procedure. Furthermore, we propose an instance selection method based on random mutation hill climbing (RMHC) algorithm with MapReduce framework, called RMHC-MR. The experimental result shows that RMHC-MR has a good performance in terms of classification accuracy and reduction rate.

Single string based global optimizer for geometry optimization in strongly coupled finite clusters: An adaptive mutation-driven strategy

We propose and implement a simple adaptive heuristic to optimize the geometries of clusters of point charges or ions with the ability to find the global minimum energy configurations. The approach uses random mutations of a single string encoding the geometry and accepts moves that decrease the energy. Mutation probability and mutation intensity are allowed to evolve adaptively on the basis of continuous evaluation of past explorations. The resulting algorithm has been called Completely Adaptive Random Mutation Hill Climbing method. We have implemented this method to search through the complex potential energy landscapes of parabolically confined 3D classical Coulomb clusters of hundreds or thousands of charges--usually found in high frequency discharge plasmas. The energy per particle (EN∕N) and its first and second differences, structural features, distribution of the oscillation frequencies of normal modes, etc., are analyzed as functions of confinement strength and the number of charges in the system. Certain magic numbers are identified. In order to test the feasibility of the algorithm in cluster geometry optimization on more complex energy landscapes, we have applied the algorithm for optimizing the geometries of MgO clusters, described by Coulomb-Born-Mayer potential and finding global minimum of some Lennard-Jones clusters. The convergence behavior of the algorithm compares favorably with those of other existing global optimizers.

Building lightweight intrusion detection system using wrapper-based feature selection mechanisms

Intrusion Detection System (IDS) is an important and necessary component in ensuring network security and protecting network resources and network infrastructures. How to build a lightweight IDS is a hot topic in network security. Moreover, feature selection is a classic research topic in data mining and it has attracted much interest from researchers in many fields such as network security, pattern recognition and data mining. In this paper, we effectively introduced feature selection methods to intrusion detection domain. We propose a wrapper-based feature selection algorithm aiming at building lightweight intrusion detection system by using modified random mutation hill climbing (RMHC) as search strategy to specify a candidate subset for evaluation, as well as using modified linear Support Vector Machines (SVMs) iterative procedure as wrapper approach to obtain the optimum feature subset. We verify the effectiveness and the feasibility of our feature selection algorithm by several experiments on KDD Cup 1999 intrusion detection dataset. The experimental results strongly show that our approach is not only able to speed up the process of selecting important features but also to yield high detection rates. Furthermore, our experimental results indicate that intrusion detection system with feature selection algorithm has better performance than that without feature selection algorithm both in detection performance and computational cost.

Finding optimal vaccination strategies for pandemic influenza using genetic algorithms

In the event of pandemic influenza, only limited supplies of vaccine may be available. We use stochastic epidemic simulations, genetic algorithms (GA), and random mutation hill climbing (RMHC) to find optimal vaccine distributions to minimize the number of illnesses or deaths in the population, given limited quantities of vaccine. Due to the non-linearity, complexity and stochasticity of the epidemic process, it is not possible to solve for optimal vaccine distributions mathematically. However, we use GA and RMHC to find near optimal vaccine distributions. We model an influenza pandemic that has age-specific illness attack rates similar to the Asian pandemic in 1957–1958 caused by influenza A(H2N2), as well as a distribution similar to the Hong Kong pandemic in 1968–1969 caused by influenza A(H3N2). We find the optimal vaccine distributions given that the number of doses is limited over the range of 10–90% of the population. While GA and RMHC work well in finding optimal vaccine distributions, GA is significantly more efficient than RMHC. We show that the optimal vaccine distribution found by GA and RMHC is up to 84% more effective than random mass vaccination in the mid range of vaccine availability. GA is generalizable to the optimization of stochastic model parameters for other infectious diseases and population structures.

Changing Representations During Search: A Comparative Study of Delta Coding

Delta coding is an iterative genetic search strategy that dynamically changes the representation of the search space in an attempt to exploit different problem representations. Delta coding sustains search by reinitializing the population at each iteration of search. This helps to avoid the asymptotic performance typically observed in genetic search as the population becomes more homogeneous. Here, the optimization ability of delta coding is empirically compared against CHC, ESGA, GENITOR, and random mutation hill-climbing (RMHC) on a suite of well-known test functions with and without Gray coding. Issues concerning the effects of Gray coding on these test functions are addressed.