CalcGraph: taming the high costs of deep learning using models

Software and Systems Modeling - Models based on differential programming, like deep neural networks, are well established in research and able to outperform manually coded counterparts in many...     

A distance measure between labeled combinatorial maps

Combinatorial maps are widely used in image representation and processing, however map matching problems have not been extensively researched. This paper addresses the problem of inexact matching between labeled combinatorial maps. First, the concept of edit distance is extended to combinatorial maps, and then used to define mapping between combinatorial maps as a sequence of edit operations that transforms one map into another. Subsequently, an optimal approach based on A^* algorithm and an approximate approach based on Greedy algorithm are proposed to compute the distance between combinatorial maps. Experimental results show that the proposed inexact map matching approach produces richer search results than the exact map matching technique by tolerating small difference between maps. The proposed approach performs better in practice than the previous approach based on maximum common submap which cannot be directly used for comparing labels on the maps.     

Structural phase transition and high temperature phase structure of Friedels salt, 3CaO · Al2O3 · CaCl2 · 10H2O

Friedels salt, the chlorinated compound 3CaO · Al₂O₃ · CaCl₂ · 10H₂O (AFm phase), presents a structural phase transition at about 30°C from a monoclinic to a rhombohedral phase. It has been studied by X-ray powder diffraction and optical microscopy in transmitted light with crossed polarisers on single crystals prepared by hydrothermal synthesis. The high temperature phase was determined at 37°C from X-ray single crystal diffraction data. The compound crystallises in the space group R c with lattice parameters of a = 5.7358(6)Å and c = 46.849(9)Å (Z = 3 and D_x = 2.111 g/cm³). The refinement of 498 independent reflections with I > 2σ(I) led to a residual factor of 7.1%. The Friedels salt can be described as a layered structure with positively charged main layers of composition [Ca₂Al(OH)₆]⁺ and negatively charged layers of composition [Cl⁻,2H₂O]. The chloride anions are surrounded by 10 hydrogen atoms, of which six belong to hydroxyl groups and four to water molecules. The structural phase transition may be related to the size of the chloride anions, which are not adapted to the octahedral cavity formed by bonded water molecules.     

ITER Hot Cell optimisation

The Hot Cell has the pivotal role in supporting on-going maintenance of the ITER machine. The experimental nature of the ITER Tokamak dictates that Hot Cell tasks will be complex in nature and will evolve over the lifetime of the project. This paper presents the results of a study commissioned by the ITER Organisation and undertaken by Oxford Technologies Ltd and Comex Nucléaire to investigate optimisation of the Hot Cell baseline design to ensure it is able to fulfil these challenging requirements.The study reviews the rationale for the current baseline Hot Cell design, derives a set of design variables which have a significant influence on the Hot Cell (e.g. Divertor refurbishment strategy) and assesses the impact of these variables. The method adopted was to perform a quantitative analysis of the impact of each variable individually without consideration of any other factors. Following on from this a multi-dimensional analysis, performed by expert assessment, was undertaken to assess the relative benefits and limitations of all of these interacting variables. The output of this study are alternative Hot Cell layouts which are optimised for criteria considered critical to the ITER Organisation, i.e. maximising Tokamak availability and minimising Hot Cell cost and size. This paper discusses the overall study process, issues raised and results.     

Novel allergen structures with tandem amino acid repeats derived from German and American cockroach

Cockroaches produce potent allergens that are an important cause of asthma. The two principal domiciliary cockroach species, Blattella germanica and Periplaneta americana, secrete major allergens, Bla g 1 and Per a 1. Here, we report the molecular cloning of three Bla g 1 cDNA clones, which showed 70% amino acid sequence identity with Per a 1. Plaque immunoassays with human IgE antibodies or murine monoclonal antibodies showed that these allergens were antigenically cross-reactive. The Bla g 1 sequences also showed homology to five previously undefined cockroach allergen sequences. An unusual feature of all these sequences was that they contained multiple tandem amino acid repeats of approximately 100 amino acid residues. Between one and seven repeat units were identified by dot-plot matrix analysis. The sequences also showed homology to a mosquito protein involved in digestion (ANG12 precursor) and to mitochondrial energy transfer proteins. High levels of Bla g 1 were found in cockroach hindgut and proventriculus. Amino acid sequencing of natural Bla g 1 and Per a 1 suggested that these allergens are cleaved by trypsin-like enzymes following secretion into the digestive tract. The repeat sequences appear to have evolved by duplication of an ancestral amino acid domain, which may have arisen from the mitochondrial energy transfer proteins.     

Overview of the KOAN programming environment for the iPSC/2 and performance evaluation of the BECAUSE test program 2.51

In this paper, we describe a Fortran programming environment using the KOAN Shared Virtual Memory on a distributed memory parallel architecture. We then discuss its use for parallelizing a benchmark application and present performance results. Initial findings show that it is possible to obtain good performance without the need for low level parallel programming.     

SPHINX: Schema integration by example

The Internet has instigated a critical need for automated tools that facilitate integrating countless databases. Since nontechnical end users are often the ultimate repositories of the domain information required to distinguish differences in data types, an effective solution must integrate simple GUI based data browsing tools and automatic mapping methods that eliminate the requirement for a technical user to supervise the process. We develop a metamodel of data integration as the basis for absorbing feedback from an end user. The schema integration algorithm draws examples from the data and learns integrating view definitions by asking a user simple yes or no questions. The metamodel enables a search mechanism that is guaranteed to converge to a correct integrating view definition without the user having to know a view definition language such as SQL or SchemaSQL, or even having to inspect the final view definition. We show how data catalog statistics, normally used to optimize queries, can be exploited to parameterize the search heuristics and improve the convergence of the learning algorithm.     

Modelling the individual cell lag phase: effect of temperature and pH on the individual cell lag distribution of Listeria monocytogenes

The individual-based approach of the lag phase is gaining interest, especially for pathogens that initially contaminate food products in low amounts. In this paper, the effect of temperature (30, 10, 7, 4 and 2 °C) and pH (7.4, 6.1, 5.5, 5.0, 4.7 and 4.4) on the individual cell lag phase of Listeria monocytogenes was examined in a factorial design, using OD measurements. Individual lag phases of about 100 individual cells per condition were examined and calculated using a linear extrapolation method. Generation times were calculated out of the slope.

The obtained data were analyzed at three different levels: in a first approach, the mean values were calculated for each set of environmental conditions and compared to predictions made by the USDA's Pathogen Modeling Program (PMP) for analogous growth conditions. The PMP predictions of the generation times were in the same order of magnitude as the obtained data, although a persistent underestimation could be observed. The observed individual cell lag data differed from lag phase predictions by PMP. Possible reasons for this discrepancy are discussed.

Secondly, histograms of individual lag phase measurements were constructed for the different temperature–pH combinations. In this way, the influence of both factors on the variability of individual lag phases could be estimated. At low stress levels, most individual cells showed a short lag phase resulting in a compression of the histograms at the zero-lag level, while, at high stress levels, the histograms shifted to longer lag phases with a significant increase in variability.

Thirdly, 37 different distribution types were fitted to the datasets to reveal the distributions that fitted best the obtained data. The gamma distribution was preferred at moderate stress levels, while the Weibull distribution was chosen for harsher growth conditions.     

Obtaining single cells: analysis and evaluation of an experimental protocol by means of a simulation model

The research presented in this paper analyses a newly developed experimental protocol for isolating single cells by constructing a simulation model of the process. The protocol involves sequential 50% dilutions of a cell suspension in a microtiter plate, so that eventually, wells are obtained containing exactly one cell.

The aim of this modelling study is (i) to gain insight in the governing mechanisms of the dilution process, (ii) to confirm experimental findings and (iii) to enable the prediction of an average outcome for future experiments.

The model construction process is presented chronologically. The initial basic model simulates the experiment as a sequence of binomial processes, using Monte Carlo techniques. Statistical analysis of the results shows that aggregational factors need to be taken into account in the form of a lognormal distribution. Several issues involved in this adaptation are discussed. To fully account for cell aggregation in the dilution process, a cell clumping algorithm is built into the simulation model. Simulation data from the resulting model show similar statistical characteristics as the experimental data and yield reliable prediction intervals for the available experimental data.

The simulation model is a useful tool to support experimental findings and predict the outcome of future experiments. Even more importantly, this study emphasises the importance of careful statistical analysis in single cell research. The impact of stochastic effects is considerably amplified at the low cell concentrations involved and needs to be taken into account in any modelling effort.