Fingerprinting: Visualization and Automatic Analysis of Prisoner's Dilemma Strategies

Fingerprinting is a technique for generating a representation-independent functional signature for a game playing agent. Fingerprints can be used to compare agents across representations in an automatic fashion. The theory of fingerprints is developed for software agents that play the iterated prisoner's dilemma. Examples of the technique for computing fingerprints are given. This paper summarizes past results and introduces the following new results. Fingerprints of prisoner's dilemma strategies that are represented as finite-state machines must be rational functions. An example of a strategy that does not have a finite-state representation and which does not have a rational fingerprint function is given: the majority strategy. It is shown that the AllD- and AllC-based fingerprints can be derived from the tit-for-tat fingerprint by a simple substitution. Fingerprints for four new probe strategies are introduced, generalizing previous work in which tit-for-tat is the sole probe strategy. A trial comparison is made of evolved prisoner's dilemma strategies across three representations: finite-state machines, feedforward neural nets, and lookup tables. Fingerprinting demonstrates that all three representations sample the strategy space in a radically different manner, even though the neural net's and lookup table's parameters are alternate encodings of the same strategy space. This space of strategies is also a subset of those encoded by the finite-state representation. Shortcomings of the fingerprint technique are outlined, with illustrative examples, and possible paths to overcome these shortcomings are given.     

Mechanistic-based comparisons of stabilised base and granular surface layers of low-volume roads

Granular surface and base layers of low-volume roads (LVRs) are frequently subjected to severe damage that adversely affects safety and requires regular repair and maintenance. Various stabilisation methods have been evaluated for mitigating damage and improving serviceability of LVR systems. However, few well-documented comparisons exist of the field mechanical performance, durability and construction costs of different stabilisation methods under the same set of geological, climate, and traffic conditions. Therefore, the present study was conducted to identify the most effective and economical among several stabilisation methods for repairing or reconstructing granular surface and base layers of LVRs. In this study, a range of promising technologies from a comprehensive literature review was selected and examined using field demonstration sections. A total of nine geomaterials, three chemical stabilisers, and three types of geosynthetics were used to construct various test sections over a 3.22 km stretch of granular-surfaced road. Extensive falling weight deflectometer (FWD) and dynamic cone penetrometer tests were performed to evaluate the multilayered elastic moduli and strengths of the various sections. This paper details the design and construction of each test section, compares the as-constructed mechanistic performance of all test sections, and assesses stiffness changes of several sections one year after construction. To provide a statistical basis for the comparisons, a pairwise multiple-comparison procedure applied for unequal sample sizes and variances and the paired t-test were used to analyse the FWD test results, demonstrating that the performance measures of the various sections were significantly different.     

A novel engineering tool for creative design of fluid systems

Computational fluid dynamics is not often used early in the conceptual design stage of product development due to the lengthy computation times involved with solving complex computational fluid dynamics models. At this early stage, design options are being explored and significant changes are common, and therefore updated solutions must be found quickly to make these models effective. Because of this, computational fluid dynamics models are often reduced to analysis tools used later in the process and are used for refinement rather than for creative engineering design. This paper presents a novel method to create computational fluid dynamics models that can be used earlier in the engineering design process. The key aspects of analysis models used in the initial, creative phase of design are the ability to make changes and re-analyze the altered model quickly. Typically, computational fluid dynamics analysts choose to re-analyze the entire altered model to maintain the same level of accuracy. This can take a significant amount of time because the entire domain must be recalculated. Much of this time is devoted to fine-tuning the model, i.e., improving the accuracy of details of the domain that are sometimes non-essential to the bulk characteristics of the flowfield. However, in the early stage of the design process, decisions are often made based on the large-scale behavior of the fluid flow; fine details are often inconsequential. We have taken advantage of this premise to decrease the turnaround time required to re-analyze a computational fluid dynamics model using the Adaptive Modeling by Evolving Blocks Algorithm. The Adaptive Modeling by Evolving Blocks Algorithm is a genetic programming-based optimization program that segregates a flowfield and places minimal cost solvers in regions with simple flow dynamics while placing full-scale computational fluid dynamics solvers in the more complex regions to preserve accuracy. The program evolves the combined segregation scheme and solver placement until a reliably accurate, faster multi-solver model is found. Substantial reductions in solution times have been found in some cases. The results show an improvement in the speed of the multi-solver when compared with a single-model solution with no significant loss of accuracy.     

Method of Adaptive-Gradient Elements for Computational Mechanics

For tackling high-gradient, localized, or singular boundary value problems, the concept of an adaptive-gradient (AG) element family is introduced to advance the utility of discretization methods. Capable of encompassing regular and singular elements as special cases, a basic but versatile family of AG elements for multidimensional applications is derived whose gradient and singularity can be controlled parametrically to handle a wide variety of functional behavior with standard mesh configurations. As illustrations, examples of usage and performance in a set of linear and nonlinear mixed-boundary value problems are presented.     

Planned tournament selection

Tournament selection is a versatile method of selection and replacement used in evolutionary computation. Normally tournaments are chosen uniformly at random. This study demonstrates the effectiveness of planning tournaments in advance to control information flow within a population being evolved for optimization. Tests are performed on a variety of evolutionary test problems, finding that different planned tournament schemes yield significant differences in performance. The correct type of planned tournament is found to be problem dependent. In addition to a linear-function scheme for planning tournaments, this study also introduces a technique called multi-deme planned tournaments selection which permits simpler cases of a problem to be automatically used to reduce the time required to solve more complex cases.     

Experimental Response of Piles in Sand under Compound Motion

This paper describes a fundamental experimental study on the vertical, horizontal, and rocking dynamic behavior of single pile foundations in granular soils. Aimed at generating an extensive experimental database with sufficient parametric variations to clarify a number of issues, multiple series of canonical small-strain forced-vibration centrifuge tests were performed on two model piles using the technique of random vibration and impact loading. Correlated well with vertical- and horizontal-centric dynamic tests within the experimental program, a novel hybrid-mode test method by means of eccentric excitation is validated and employed for the characterization of the foundation responses in general planar motion. A large set of experimental data for different length scales were generated and synthesized in the frequency domain in the form of directional force-response transfer functions. By virtue of the physical measurements, the validity and limitations of two fundamental elastodynamic pile solutions pertaining to the physical dynamic soil-foundation problem are also evaluated.     

Human artificial chromosomes generated by modification of a yeast artificial chromosome containing both human alpha satellite and single-copy DNA sequences

A human artificial chromosome (HAC) vector was constructed from a 1-Mb yeast artificial chromosome (YAC) that was selected based on its size from among several YACs identified by screening a randomly chosen subset of the Centre d'Etude du Polymorphisme Humain (CEPH) (Paris) YAC library with a degenerate alpha satellite probe. This YAC, which also included non-alpha satellite DNA, was modified to contain human telomeric DNA and a putative origin of replication from the human beta-globin locus. The resultant HAC vector was introduced into human cells by lipid-mediated DNA transfection, and HACs were identified that bound the active kinetochore protein CENP-E and were mitotically stable in the absence of selection for at least 100 generations. Microdissected HACs used as fluorescence in situ hybridization probes localized to the HAC itself and not to the arms of any endogenous human chromosomes, suggesting that the HAC was not formed by telomere fragmentation. Our ability to manipulate the HAC vector by recombinant genetic methods should allow us to further define the elements necessary for mammalian chromosome function.     

Tensionless contact of a flexible plate and annulus with a smooth half-space under axisymmetric loads by integral equations

In this paper, a new hybrid method of analysis is presented for the problem of a frictionless circular plate or annulus in tensionless contact with a half-space. By virtue of a set of analytically explicit Green's functions for the two interacting continua, an exact but compact integral equation formulation with closed-form kernels is derived. With the incorporation of a newly developed adaptive-gradient (AG) element capable of capturing regular-to-singular solution transitions smoothly, an accurate numerical procedure is developed and validated in a number of benchmark cases of nonlinear plate–annulus–half-space interaction. From the simplicity and rate of convergence demonstrated, the hybrid method is apt to be an attractive analytical–numerical platform that can be extended to a large class of contact problems.     

Comprehensive bidding strategies with genetic programming/finitestate automata

This research is an extension of the authors' previous work in double auctions aimed at developing bidding strategies for electric utilities which trade electricity competitively. The improvements detailed in this paper come from using data structures which combine genetic programming and finite state automata termed GP-Automata. The strategies developed by the method described here are adaptive-reacting to inputs-whereas the previously developed strategies were only suitable in the particular scenario for which they had been designed. The strategies encoded in the GP-Automata are tested in an auction simulator. The simulator pits them against other distribution companies (distcos) and generation companies (gencos), buying and selling power via double auctions implemented in regional commodity exchanges. The GP-Automata are evolved with a genetic algorithm so that they possess certain characteristics. In addition to designing successful bidding strategies (whose usage would result in higher profits) the resulting strategies can also be designed to imitate certain types of trading behaviors. The resulting strategies can be implemented directly in online trading, or can be used as realistic competitors in an off-line trading simulator