General-purpose Implementation Research Articles

Gradient boosting for convex cone predict and optimize problems

Prediction models are traditionally optimized independently from decision-based optimization. Conversely, a ‘smart predict then optimize’ (SPO) framework optimizes prediction models to minimize downstream decision regret. In this paper we present dboost, the first general purpose implementation of smart gradient boosting for ‘predict, then optimize’ problems. The framework supports convex quadratic cone programming and gradient boosting is performed by implicit differentiation of a custom fixed-point mapping. Experiments comparing with state-of-the-art SPO methods show that dboost can further reduce out-of-sample decision regret.

BayesicFitting, a PYTHON toolbox for Bayesian fitting and evidence calculation.

BayesicFitting is a comprehensive, general-purpose toolbox for simple and standardized model fitting. Its fitting options range from simple least-squares methods, via maximum likelihood to fully Bayesian inference, working on a multitude of available models. BayesicFitting is open source and has been in development and use since the 1990s. It has been applied to a variety of science applications, chiefly in astronomy. BayesicFitting consists of a collection of PYTHON classes that can be combined to solve quite complicated inference problems. Amongst the classes are models, fitters, priors, error distributions, engines, samples, and of course NestedSampler, our general-purpose implementation of the nested sampling algorithm. Nested sampling is a novel way to perform Bayesian calculations. It can be applied to inference problems, that consist of a parameterized model to fit measured data to. NestedSampler calculates the Bayesian evidence as the numeric integral over the posterior probability of (hyper)parameters of the problem. The solution in terms of the parameters is obtained as a set of weighted samples drawn from the posterior. In this paper, we emphasize nested sampling and all classes that are directly connected to it. Additionally, we present the fitters, which fit the data by the least-squares method or the maximum likelihood method. They can also calculate the Bayesian evidence as a Gaussian approximation. We will discuss the architecture of the toolbox. Which classes are present, what is their function, how they are related and implementational details where it gets complicated. • Python toolbox for generic model fitting. • Easy usage and open source. • Ready-made combinable models. • Least squares, maximum likelihood and Nested Sampling. • Evidence calculation.

Revisiting high-order Taylor methods for astrodynamics and celestial mechanics

ABSTRACT We present heyoka, a new, modern and general-purpose implementation of Taylor’s integration method for the numerical solution of ordinary differential equations. Detailed numerical tests focused on difficult high-precision gravitational problems in astrodynamics and celestial mechanics show how our general-purpose integrator is competitive with and often superior to state-of-the-art specialized symplectic and non-symplectic integrators in both speed and accuracy. In particular, we show how Taylor methods are capable of satisfying Brouwer’s law for the conservation of energy in long-term integrations of planetary systems over billions of dynamical time-scales. We also show how close encounters are modelled accurately during simulations of the formation of the Kirkwood gaps and of Apophis’ 2029 close encounter with the Earth (where heyoka surpasses the speed and accuracy of domain-specific methods). heyoka can be used from both C++ and python, and it is publicly available as an open-source project.

PorE: A code for deterministic and systematic analyses of porosities.

Accurate numerical calculations of porosities and related properties are of importance when analyzing metal-organic frameworks (MOFs). We present porE, an open-source, general-purpose implementation to compute such properties and discuss all results regarding their sensitivity to numerical parameters. Our code combines the numerical efficiency of Fortran with the user-friendliness of Python. Three different approaches to calculate porosities are implemented in porE, and their advantages and drawbacks are discussed. In contrast to commonly used implementations, our approaches are entirely deterministic and do not require any stochastic averaging. In addition to the calculation of porosities, porE can calculate pore size distributions and offers the possibility to analyze pore windows. The underlying approaches are outlined, and pore windows are discussed concerning their impact on the analyzed porosities. Comparisons with reference values aim for a clear differentiation between void and accessible porosities, which we provide for a small benchmark set consisting of eight MOFs. In addition, our approaches are used for a bigger benchmark set containing 370 MOFs, where we determine linear relationships within our approaches as well as to reference values. We show how these relationships can be used to derive corrections to a give porosity approach, minimizing its mean error. As a highlight we show how complex workflows can be designed with a few lines of Python code using porE.

Smart Helmet: Thresh-Learner–Online Machine Learning on Data Streams

Today, with an enormous generation and availability of time series data and streaming data, there is an increasing need for an automatic analyzing architecture to get fast interpretations and results. One of the significant potentiality of streaming analytics is to train and model each stream with unsupervised Machine Learning (ML) algorithms to detect anomalous behaviors, fuzzy patterns, and accidents in real-time. If executed reliably, each anomaly detection can be highly valuable for the application. In this paper, we propose a dynamic threshold setting system denoted as Thresh-Learner, mainly for the Internet of Things (IoT) applications that require anomaly detection. The proposed model enables a wide range of real-life applications where there is a necessity to set up a dynamic threshold over the streaming data to avoid anomalies, accidents or sending alerts to distant monitoring stations.We took the major problem of anomalies and accidents in coal mines due to coal fires and explosions. This results in loss of life due to the lack of automated alarming systems. We propose Thresh-Learner, a general purpose implementation for setting dynamic thresholds. We illustrate it through the Smart Helmet for coal mine workers which seamlessly integrates monitoring, analyzing and dynamic thresholds using IoT and analysis on the cloud.

Partial Compilation of ASP Programs

AbstractAnswer Set Programming (ASP) is a well-known declarative formalism in logic programming. Efficient implementations made it possible to apply ASP in many scenarios, ranging from deductive databases applications to the solution of hard combinatorial problems. State-of-the-art ASP systems are based on the traditional ground&solve approach and are general-purpose implementations, i.e., they are essentially built once for any kind of input program. In this paper, we propose an extended architecture for ASP systems, in which parts of the input program are compiled into an ad-hoc evaluation algorithm (i.e., we obtain a specific binary for a given program), and might not be subject to the grounding step. To this end, we identify a condition that allows the compilation of a sub-program, and present the related partial compilation technique. Importantly, we have implemented the new approach on top of a well-known ASP solver and conducted an experimental analysis on publicly-available benchmarks. Results show that our compilation-based approach improves on the state of the art in various scenarios, including cases in which the input program is stratified or the grounding blow-up makes the evaluation unpractical with traditional ASP systems.

GPGPU Virtualization Techniques a Comparative Survey

The Graphic Processing Units (GPU) are being adopted in many High Processing Computing (HPC) facilities because of their massively parallel and extraordinary computing power, which makes it possible to accelerate many general purpose implementations from different domains. A general-purpose GPU (GPGPU) is a GPU that performs computations that were traditionally handled by central processing unit (CPU) to accelerate applications along with handling traditional computations for graphics rendering. However, GPUs have some limitations, such as increased acquisition costs as well as larger space requirements, more powerful energy supplies, and their utilization is usually low for most workloads. That results in the need of GPU virtualization to maximize the use of an acquired GPU to share between the virtual machines for optimal use, to reduce power utilization and minimize the costs. This study comparatively reviews the recent GPU virtualization techniques including API remoting, para, full and hardware based virtualization, targeted for general-purpose accelerations.

Solutions for General-Purpose Electromagnetic Problems Using the Random Auxiliary Sources Method

A 3-D general-purpose implementation of the random auxiliary sources (RASs) method is proposed, benefiting from the Rao–Wilton–Glisson (RWG) testing functions. The testing procedure is presented in terms of the RWG function parameters. The performance of the proposed RAS with the RWG testing function is compared with point matching. Also, a simple treatment for mixed boundary problems is introduced to enable solving the complex electromagnetic problems. Furthermore, a waveport formulation is proposed that is better suited for the iterative approach of the RAS method to facilitate the modular analysis of components. A gradient-based testing formulation for the transmitting port is exploited to enforce the matching to the forward wave only on a waveport. Nevertheless, several test cases are presented to evaluate the accuracy and performance of the proposed method in comparison to the commercial full-wave solvers showing a reasonable agreement.

Efficient automatic discrete adjoint sensitivity computation for topology optimization – heat conduction applications

PurposeTopology optimization is a method used for developing optimized geometric designs by distributing material pixels in a given design space that maximizes a chosen quantity of interest (QoI) subject to constraints. The purpose of this study is to develop a problem-agnostic automatic differentiation (AD) framework to compute sensitivities of the QoI required for density distribution-based topology optimization in an unstructured co-located cell-centered finite volume framework. Using this AD framework, the authors develop and demonstrate the topology optimization procedure for multi-dimensional steady-state heat conduction problems.Design/methodology/approachTopology optimization is performed using the well-established solid isotropic material with penalization approach. The method of moving asymptotes, a gradient-based optimization algorithm, is used to perform the optimization. The sensitivities of the QoI with respect to design variables, required for optimization algorithm, are computed using a discrete adjoint method with a novel AD library named residual automatic partial differentiator (Rapid).FindingsTopologies that maximize or minimize relevant quantities of interest in heat conduction applications are presented. The efficacy of the technique is demonstrated using a variety of realistic heat transfer applications in both two and three dimensions, in conjugate heat transfer problems with finite conductivity ratios and in non-rectangular/non-cuboidal domains.Originality/valueIn contrast to most published work which has either used finite element methods or Cartesian finite volume methods for transport applications, the topology optimization procedure is developed in a general unstructured finite volume framework. This permits topology optimization for flow and heat transfer applications in complex design domains such as those encountered in industry. In addition, the Rapid library is designed to provide a problem-agnostic pathway to automatically compute all required derivatives to machine accuracy. This obviates the necessity to write new code for finding sensitivities when new physics are added or new cost functions are considered and permits general-purpose implementations of topology optimization for complex industrial applications.

Parallel Fokker–Planck-DSMC algorithm for rarefied gas flow simulation in complex domains at all Knudsen numbers

A major challenge for the conventional Direct Simulation Monte Carlo (DSMC) technique lies in the fact that its computational cost becomes prohibitive in the near continuum regime, where the Knudsen number (Kn)—characterizing the degree of rarefaction—becomes small. In contrast, the Fokker–Planck (FP) based particle Monte Carlo scheme allows for computationally efficient simulations of rarefied gas flows in the low and intermediate Kn regime. The Fokker–Planck collision operator—instead of performing binary collisions employed by the DSMC method—integrates continuous stochastic processes for the phase space evolution in time. This allows for time step and grid cell sizes larger than the respective collisional scales required by DSMC. Dynamically switching between the FP and the DSMC collision operators in each computational cell is the basis of the combined FP-DSMC method, which has been proven successful in simulating flows covering the whole Kn range. Until recently, this algorithm had only been applied to two-dimensional test cases. In this contribution, we present the first general purpose implementation of the combined FP-DSMC method. Utilizing both shared- and distributed-memory parallelization, this implementation provides the capability for simulations involving many particles and complex geometries by exploiting state of the art computer cluster technologies.

Second order accurate particle-based formulations

Purpose– The purpose of this paper is to present new implementation aspects of unified explicit time integration algorithms, called the explicit GS4-II family of algorithms, of a second-order time accuracy in all the unknowns (e.g. positions, velocities, and accelerations) with particular attention to the moving-particle simulation (MPS) method for solving the incompressible fluids with free surfaces.Design/methodology/approach– In the present paper, the explicit GS4-II family of algorithms is implemented in the MPS method in the following two different approaches: a direct explicit formulation with the use of the weak incompressibility equation involving the (modified) speed of sound; and a predictor-corrector explicit formulation. The first approach basically follows the concept of the explicit MPS method, presented in the literature, and the latter approach employs a similar concept used in, for example, a fractional-step method in computational fluid dynamics.Findings– Illustrative numerical examples demonstrate that any scheme within the proposed algorithmic framework captures the physics with the necessary second-order time accuracy and stability.Originality/value– The new algorithmic framework extended with the GS4-II family encompasses a multitude of pastand new schemes and offers a general purpose and unified implementation.

Embedded Coprocessors for Native Execution of Geometric Algebra Operations

Clifford algebra or geometric algebra (GA) is a simple and intuitive way to model geometric objects and their transformations. Operating in high-dimensional vector spaces with significant computational costs, the practical use of GA requires dedicated software and/or hardware architectures to directly support Clifford data types and operators. In this paper, a family of embedded coprocessors for the native execution of GA operations is presented. The paper shows the evolution of the coprocessor family focusing on the latest two architectures that offer direct hardware support to up to five-dimensional Clifford operations. The proposed coprocessors exploit hardware-oriented representations of GA elements and operators properly conceived to obtain fast performing implementations. The coprocessor prototypes, implemented on field programmable gate arrays development boards, show significant speedups of about one order of magnitude with respect to the baseline software library Gaigen running on a general-purpose processor. The paper also presents an execution analysis of different GA-based applications, namely inverse kinematics of a robot, optical motion capture, raytracing, and medical image processing, showing good speedups with respect to the baseline general-purpose implementation.

Guard-based partial-order reduction

This paper aims at making partial-order reduction independent of the modeling language. To this end, we present a guard-based method which is a general-purpose implementation of the stubborn set method. We approach the implementation through so-called necessary enabling sets and do-not-accord sets, and give an algorithm suitable for an abstract model checking interface. We also introduce necessary disabling sets and heuristics to produce smaller stubborn sets and thus better reduction at low costs. We explore the effect of these methods using an implementation in the model checker LTSmin. We experiment with partial-order reduction on a number of Promela models, on benchmarks from the BEEM database in the DVE language, and with several with LTL properties. The efficiency of the heuristic algorithm is established by a comparison to the subset-minimal Deletion algorithm and the simple closure algorithm. We also compare our results to the Spin model checker. While the reductions take longer, they are consistently better than Spin 's ample set and often surpass the upper bound for the process-based ample sets, established empirically earlier on BEEM models.

Linux Support for Fast Transparent General Purpose Checkpoint/Restart of Multithreaded Processes in Loadable Kernel Module

Checkpoint/Restart is the ability to save the state of a running application so that it can later resume its execution from the time of the checkpoint. These are techniques with many potential applications, including establishment of a fault-tolerant environment, improving system resource utilization, and true migration of a process. With increasing hardware speed and size of clusters the average time between failures has been reduced. Therefore, fault tolerance and ability to checkpoint a process have become inevitable. Almost all platforms deployed for high-performance computing support process checkpoint/restart. Linux as one of the popular operating systems does not provide a general purpose implementation. Some are limited to specific type of parallel programming library, confined to some unique well-behaved type of applications, or reliant on specific features in kernel which could be missing on many occasions. Most of implementations demand elaborate practice of recompiling a whole kernel to apply required patches. In this paper, we describe the design and implementation of multithreaded process checkpoint/restart system for Linux which provide capability of dynamic extension to increase compatibility and reduce system overhead. It does not impose any requirement on the existence of a special facility in the operating system and can do checkpoint/restart of an application independent of their behavior and fully transparent. The entire system is absolutely implemented in multiple kernel loadable modules, which result in ease of use and eliminate the burden of complex system administration.

On the implementation ofGNU Prolog

AbstractGNU Prologis a general-purpose implementation of the Prolog language, which distinguishes itself from most other systems by being, above all else, a native-code compiler which produces stand-alone executables which do not rely on any bytecode emulator or meta-interpreter. Other aspects which stand out include the explicit organization of the Prolog system as a multipass compiler, where intermediate representations are materialized, in Unix compiler tradition.GNU Prologalso includes an extensible and high-performance finite-domain constraint solver, integrated with the Prolog language but implemented using independent lower-level mechanisms. This paper discusses the main issues involved in designing and implementingGNU Prolog: requirements, system organization, performance, and portability issues as well as its position with respect to other Prolog system implementations and the ISO standardization initiative.

DFTB+ and lanthanides

DFTB+ is a recent general purpose implementation of density-functional based tight binding. One of the early motivators to develop this code was to investigate lanthanide impurities in nitride semiconductors, leading to a series of successful studies into structure and electrical properties of these systems. Here we describe our general framework to treat the physical effects needed for these problematic impurities within a tight-binding formalism, additionally discussing forces and stresses in DFTB. We also present an approach to evaluate the general case of Slater-Koster transforms and all of their derivatives in Cartesian coordinates. These developments are illustrated by simulating isolated Gd impurities in GaN.

Pervasive pheromone-based interaction with RFID tags

Despite the growing interest in pheromone-based interaction to enforce adaptive and context-aware coordination, the number of deployed systems exploiting digital pheromones to coordinate the activities of situated autonomous agents is still very limited. In this article, we present a simple low-cost and general-purpose implementation of a pheromone-based interaction mechanism for pervasive environments. This is realized by making use of RFID tags to store digital pheromones and by having humans or robots spread/sense pheromones by properly writing/reading RFID tags populating the surrounding physical environment. We exemplify and evaluate the effectiveness of our approach via an application for object-tracking. This application allows robots and humans to find forgotten-somewhere objects by following pheromones trails associated with them. In addition, we sketch further potential applications of our approach in pervasive computing scenarios, discuss related work in the area, and identify future research directions.

Efficient formulation and computer implementation of the active-space electron-attached and ionized equation-of-motion coupled-cluster methods

The efficient, general-purpose implementations of the active-space electron-attached (EA) and ionized (IP) equation-of-motion coupled-cluster (EOMCC) methods including up to 3p-2h and 3h-2p excitations, called EA-EOMCCSDt and IP-EOMCCSDt, respectively, are discussed. The details of the algorithm that enables one to achieve a high degree of code vectorization for the active-space methods and the factorized forms of the EA- and IP-EOMCCSDt equations that maximize the benefits of using active orbitals in the process of selecting the dominant 3p-2h and 3h-2p excitations are presented. The results of benchmark calculations for the low-lying doublet and quartet states of the CH and SH radicals reveal that the active-space EA-EOMCCSDt and IP-EOMCCSDt methods are capable of producing results for the electronic excitations in open-shell systems that match the high accuracy of EA- and IP-EOMCC calculations with a full treatment of 3p-2h and 3h-2p excitations, even when the excited states of interest display a manifestly multideterminantal nature, with the costs that can be on the same order of those characterizing the basic EOMCC singles and doubles approach.

SIMULATED MAXIMUM LIKELIHOOD APPLIED TO NON‐GAUSSIAN AND NONLINEAR MIXED EFFECTS AND STATE–SPACE MODELS

SummaryThe paper presents an overview of maximum likelihood estimation using simulated likelihood, including the use of antithetic variables and evaluation of the simulation error of the resulting estimates. It gives a general purpose implementation of simulated maximum likelihood and uses it to re‐visit four models that have previously appeared in the published literature: a state–space model for count data; a nested random effects model for binomial data; a nonlinear growth model with crossed random effects; and a crossed random effects model for binary salamander‐mating data. In the case of the last three examples, this appears to be the first time that maximum likelihood fits of these models have been presented.

Algorithms for variable length Markov chain modeling

We present a general purpose implementation of variable length Markov models. Contrary to fixed order Markov models, these models are not restricted to a predefined uniform depth. Rather, by examining the training data, a model is constructed that fits higher order Markov dependencies where such contexts exist, while using lower order Markov dependencies elsewhere. As both theoretical and experimental results show, these models are capable of capturing rich signals from a modest amount of training data, without the use of hidden states. The source code is freely available at http://www.soe.ucsc.edu/~jill/src/