Parallel evolution using multi-chromosome cartesian genetic programming

Parallel and distributed methods for evolutionary algorithms have concentrated on maintaining multiple populations of genotypes, where each genotype in a population encodes a potential solution to the problem. In this paper, we investigate the parallelisation of the genotype itself into a collection of independent chromosomes which can be evaluated in parallel. We call this multi-chromosomal evolution (MCE). We test this approach using Cartesian Genetic Programming and apply MCE to a series of digital circuit design problems to compare the efficacy of MCE with a conventional single chromosome approach (SCE). MCE can be readily used for many digital circuits because they have multiple outputs. In MCE, an independent chromosome is assigned to each output. When we compare MCE with SCE we find that MCE allows us to evolve solutions much faster. In addition, in some cases we were able to evolve solutions with MCE that we unable to with SCE. In a case-study, we investigate how MCE can be applied to to a single objective problem in the domain of image classification, namely, the classification of breast X-rays for cancer. To apply MCE to this problem, we identify regions of interest (RoI) from the mammograms, divide the RoI into a collection of sub-images and use a chromosome to classify each sub-image. This problem allows us to evaluate various evolutionary mutation operators which can pairwise swap chromosomes either randomly or topographically or reuse chromosomes in place of other chromosomes.     

Derivation of Strictly Stable High Order Difference Approximations for Variable-Coefficient PDE

A new way of deriving strictly stable high order difference operators for partial differential equations (PDE) is demonstrated for the 1D convection diffusion equation with variable coefficients. The derivation is based on a diffusion term in conservative, i.e. self-adjoint, form. Fourth order accurate difference operators are constructed by mass lumping Galerkin finite element methods so that an explicit method is achieved. The analysis of the operators is confirmed by numerical tests. The operators can be extended to multi dimensions, as we demonstrate for a 2D example. The discretizations are also relevant for the Navier–Stokes equations and other initial boundary value problems that involve up to second derivatives with variable coefficients.     

AI and robotics in the European restaurant sector: Assessing potentials for process innovation in a high-contact service industry

Electronic Markets - The restaurant technology market is rapidly evolving and is transforming the restaurant business as a significant sector of tourism and hospitality. Enabled by artificial...     

Determination of flux directions by thermodynamic network analysis: Computing informative metabolite pools

Network thermodynamics focuses on the energetic analysis of complex metabolic networks. The method connects free Gibbs energies, metabolite concentrations and flux directions by fundamental thermodynamic laws. Here, a new application of network thermodynamics is presented that identifies those metabolite pools that have to be measured in order to determine as many flux directions as possible. For a medium-scaled reaction network such informative metabolite pools are computed with an approach based on Monte Carlo sampling. It turns out that some reactions can be directed with only a few measurements whereas other reactions cannot be directed even with a complete data set. High connectivity in metabolic reaction networks in alliance with concentration ranges make it impossible to intuitively foresee such results. In particular, the impact of measurements of a special type of metabolites being involved in many reactions, so called energetic currency metabolites is investigated.     

The influence of text modality on learning with static and dynamic visualizations

In this study we investigated the influence of text modality on learning with static and dynamic visualizations in a dynamic domain, namely the physical principles underlying fish locomotion. A 2 × 2-design with type of visualization (static vs. dynamic) and text modality (spoken vs. written) as independent variables was used. Concerning learning outcomes, it was hypothesized that (1) learners presented with dynamic visualizations would outperform learners presented with static visualizations, (2) learners presented with spoken text would outperform learners presented with written text, and (3) an interaction between type of visualization and modality would occur: the superiority of dynamic over static visualizations was expected to be more pronounced for spoken compared to written text. Subjective cognitive load measures were assessed and expected to mirror the aforementioned pattern of learning outcomes in accordance with Cognitive Load Theory (i.e., higher extraneous cognitive load (ECL) related to lower learning outcomes). For transfer tasks, the first two hypotheses could be confirmed. However, there was no interaction. Moreover, ECL was rated higher by subjects when learning with static compared to dynamic visualizations, but there were no differences for ECL with respect to the text modality. The results are discussed within the framework of Cognitive Load Theory.     

Mechanical material characterization of photosensitive polymers

In addition to the development of materials and technologies for smart HARMST applications the analysis of the thermo-mechanical material properties seems to be necessary to describe the photosensitive polymers applied. To do this, a microtensile analysis system has been developed to determine stress–strain curves, Young’s modulus and Poisson’s ratio. The results are compared with those of dynamical mechanical thermal analysis and discussed.     

Human Intelligence in the Process of Semantic Content Creation

Despite significant progress over the last years the large-scale adoption of semantic technologies is still to come. One of the reasons for this state of affairs is assumed to be the lack of useful semantic content, a prerequisite for almost every IT system or application using semantics. Through its very nature, this content can not be created fully automatically, but requires, to a certain degree, human contribution. The interest of Internet users in semantics, and in particular in creating semantic content, is, however, low. This is understandable if we think of several characteristics exposed by many of the most prominent semantic technologies, and the applications thereof. One of these characteristics is the high barrier of entry imposed. Interacting with semantic technologies today requires specific skills and expertise on subjects which are not part of the mainstream IT knowledge portfolio. A second characteristic are the incentives that are largely missing in the design of most semantic applications. The benefits of using machine-understandable content are in most applications fully decoupled from the effort of creating and maintaining this content. In other words, users do not have a motivation to contribute to the process. Initiatives in the areas of the Social Semantic Web acknowledged this problem, and identified mechanisms to motivate users to dedicate more of their time and resources to participate in the semantic content creation process. Still, even if incentives are theoretically in place, available human labor is limited and must only be used for those tasks that are heavily dependent on human intervention, and cannot be reliably automated. In this article, we concentrate on this step in between. As a first contribution, we analyze the process of semantic content creation in order to identify those tasks that are inherently human-driven. When building semantic applications involving these specific tasks, one has to install incentive schemes that are likely to encourage users to perform exactly these tasks that crucially rely on manual input. As a second contribution of the article, we propose incentives or incentive-driven tools that can be used to increase user interest in semantic content creation tasks. We hope that our findings will be adopted as recommendations for establishing a fundamentally new form of design of semantic applications by the semantic technologies community.