Fast automatic two-stage nonlinear model identification based on the extreme learning machine

It is convenient and effective to solve nonlinear problems with a model that has a linear-in-the-parameters (LITP) structure. However, the nonlinear parameters (e.g. the width of Gaussian function) of each model term needs to be pre-determined either from expert experience or through exhaustive search. An alternative approach is to optimize them by a gradient-based technique (e.g. Newton’s method). Unfortunately, all of these methods still need a lot of computations. Recently, the extreme learning machine (ELM) has shown its advantages in terms of fast learning from data, but the sparsity of the constructed model cannot be guaranteed. This paper proposes a novel algorithm for automatic construction of a nonlinear system model based on the extreme learning machine. This is achieved by effectively integrating the ELM and leave-one-out (LOO) cross validation with our two-stage stepwise construction procedure [1]. The main objective is to improve the compactness and generalization capability of the model constructed by the ELM method. Numerical analysis shows that the proposed algorithm only involves about half of the computation of orthogonal least squares (OLS) based method. Simulation examples are included to confirm the efficacy and superiority of the proposed technique.

Integrin genetic variants and stage-specific tumor recurrence in patients with stage II and III colon cancer

Integrins (ITGs) are key elements in cancer biology, regulating tumor growth, angiogenesis and lymphangiogenesis through interactions of the tumor cells with the microenvironment. Moving from the hypothesis that ITGs could have different effects in stage II and III colon cancer, we tested whether a comprehensive panel of germline single-nucleotide polymorphisms (SNPs) in ITG genes could predict stage-specific time to tumor recurrence (TTR). A total of 234 patients treated with 5-fluorouracil-based chemotherapy at the University of Southern California were included in this study. Whole-blood samples were analyzed for germline SNPs in ITG genes using PCR-restriction fragment length polymorphism or direct DNA sequencing. In the multivariable analysis, stage II colon cancer patients with at least one G allele for ITGB3 rs4642 had higher risk of recurrence (hazard ratio (HR)=4.027, 95% confidence interval (95% CI) 1.556-10.421, P=0.004). This association was also significant in the combined stage II-III cohort (HR=1.975, 95% CI 1.194-3.269, P=0.008). The predominant role of ITGB3 rs4642 in stage II diseases was confirmed using recursive partitioning, showing that ITGB3 rs4642 was the most important factor in stage II diseases. In contrast, in stage III diseases the combined analysis of ITGB1 rs2298141 and ITGA4 rs7562325 allowed to identify three distinct prognostic subgroups (P=0.009). The interaction between stage and the combined ITGB1 rs2298141 and ITGA4 rs7562325 on TTR was significant (P=0.025). This study identifies germline polymorphisms in ITG genes as independent stage-specific prognostic markers for stage II and III colon cancer. These data may help to select subgroups of patients who may benefit from ITG-targeted treatments.

Deuterium in the Galactic Centre as a result of recent infall of low-metallicity gas

The Galactic Centre is the most active and heavily processed region of the Milky Way, so it can be used as a stringent test for the abundance of deuterium (a sensitive indicator of conditions in the first 1,000 seconds in the life of the Universe). As deuterium is destroyed in stellar interiors, chemical evolution models 1 predict that its Galactic Centre abundance relative to hydrogen is D/H = 5 x 10(-12), unless there is a continuous source of deuterium from relatively primordial (low-metallicity) gas. Here we report the detection of deuterium (in the molecule DCN) in a molecular cloud only 10 parsecs from the Galactic Centre. Our data, when combined with a model of molecular abundances, indicate that D/H = (1.7 +/- 0.3) x 10(-6), five orders of magnitude larger than the predictions of evolutionary models with no continuous source of deuterium. The most probable explanation is recent infall of relatively unprocessed metal-poor gas into the Galactic Centre (at the rate inferred by Wakker(2)). Our measured D/H is nine times less than the local interstellar value, and the lowest D/H observed in the Galaxy. We conclude that the observed Galactic Centre deuterium is cosmological, with an abundance reduced by stellar processing and mixing, and that there is no significant Galactic source of deuterium.

Genome-wide association study identifies five new schizophrenia loci

We examined the role of common genetic variation in schizophrenia in a genome-wide association study of substantial size: a stage 1 discovery sample of 21,856 individuals of European ancestry and a stage 2 replication sample of 29,839 independent subjects. The combined stage 1 and 2 analysis yielded genome-wide significant associations with schizophrenia for seven loci, five of which are new (1p21.3, 2q32.3, 8p23.2, 8q21.3 and 10q24.32-q24.33) and two of which have been previously implicated (6p21.32-p22.1 and 18q21.2). The strongest new finding (P = 1.6 x 10(-11)) was with rs1625579 within an intron of a putative primary transcript for MIR137 (microRNA 137), a known regulator of neuronal development. Four other schizophrenia loci achieving genome-wide significance contain predicted targets of MIR137, suggesting MIR137-mediated dysregulation as a previously unknown etiologic mechanism in schizophrenia. In a joint analysis with a bipolar disorder sample (16,374 affected individuals and 14,044 controls), three loci reached genome-wide significance: CACNA1C (rs4765905, P = 7.0 x 10(-9)), ANK3 (rs10994359, P = 2.5 x 10(-8)) and the ITIH3-ITIH4 region (rs2239547, P = 7.8 x 10(-9)).

Energy levels, wavefunction compositions and electric dipole transitions in neutral Ca

A configuration-interaction approach, based on the use of B-spline basis sets combined with a model potential including monoelectronic and dielectronic core polarization effects, is employed to calculate term energies and wavefunctions for neutral Ca. Results are reported for singlet and triplet bound states, and some quasi-bound states above the lowest ionization limit, with angular momentum up to L = 4. Comparison with experiment and with other theoretical results shows that this method yields the most accurate energy values for neutral Ca obtained to date. Wavefunction compositions, necessary for labelling the levels, and the effects of semi-empirical polarization potentials on the wavefunctions are discussed, as are some recent identifications of doubly-excited states. It is shown that taking into account dielectronic core polarization changes the energies of the lowest terms in Ca significantly, in general by a few hundred cm(-1), the effect decreasing rapidly for the higher bound states. For Rydberg states with n approximate to 7 the accuracy of the results is often better than a few cm(-1). For series members (or perturbers) with a pronounced 3d character the error can reach 150 cm(-1). The wavefunctions are used to calculate oscillator strengths and lifetimes for a number of terms and these are compared with existing measurements. The agreement is good but points to a need for improved measurements.

It's not that I don't care, I just don't care very much: confounding between attribute non-attendance and taste heterogeneity

With the growing interest in the topic of attribute non-attendance, there is now widespread use of latent class (LC) structures aimed at capturing such behaviour, across a number of different fields. Specifically, these studies rely on a confirmatory LC model, using two separate values for each coefficient, one of which is fixed to zero while the other is estimated, and then use the obtained class probabilities as an indication of the degree of attribute non-attendance. In the present paper, we argue that this approach is in fact misguided, and that the results are likely to be affected by confounding with regular taste heterogeneity. We contrast the confirmatory model with an exploratory LC structure in which the values in both classes are estimated. We also put forward a combined latent class mixed logit model (LC-MMNL) which allows jointly for attribute non-attendance and for continuous taste heterogeneity. Across three separate case studies, the exploratory LC model clearly rejects the confirmatory LC approach and suggests that rates of non-attendance may be much lower than what is suggested by the standard model, or even zero. The combined LC-MMNL model similarly produces significant improvements in model fit, along with substantial reductions in the implied rate of attribute non-attendance, in some cases even eliminating the phenomena across the sample population. Our results thus call for a reappraisal of the large body of recent work that has implied high rates of attribute non-attendance for some attributes. Finally, we also highlight a number of general issues with attribute non-attendance, in particular relating to the computation of willingness to pay measures.

Increased anti-tumour efficacy of doxorubicin when combined with sulindac in a xenograft model of an MRP-1-positive human lung cancer

Background: A number of cellular proteins, including P-glycoprotein (P-gp) and Multiple drug Resistance Protein (MRP-1), act as drug efflux pumps and are important in the resistance of many cancers to chemotherapy. We previously reported that a small number of NSAIDs could inhibit the activity of MRP-1. Materials and Methods: We chose sulindac as a candidate agent for further investigation as it has the most favourable efficacy and toxicity profile of the agents available for a potential specific MRP-1 inhibitor. NCI H460 cells expressed MRP-1 protein (by Western blot) and also the toxicity, of doxorubicin (a substrate of MRP-1) could be potentiated in this line using non-toxic concentrations of the MRP-1 substrate/inhibitor sulindac. These cells were implanted in nude mice and the animals divided into various groups which were administered doxorubicin and/or sulindac. Results: Sulindac was shown to significantly potentiate the tumour growth inhibitor activity of doxorubicin in this MRP-1-overexpressing human tumour xenograft model. Conclusion: Sulindac may be clinically useful as an inhibitor of the MRP-1 cancer resistance mechanism.

Model selection for prognostic time-to-event gene signature discovery with applications in early breast cancer data

Model selection between competing models is a key consideration in the discovery of prognostic multigene signatures. The use of appropriate statistical performance measures as well as verification of biological significance of the signatures is imperative to maximise the chance of external validation of the generated signatures. Current approaches in time-to-event studies often use only a single measure of performance in model selection, such as logrank test p-values, or dichotomise the follow-up times at some phase of the study to facilitate signature discovery. In this study we improve the prognostic signature discovery process through the application of the multivariate partial Cox model combined with the concordance index, hazard ratio of predictions, independence from available clinical covariates and biological enrichment as measures of signature performance. The proposed framework was applied to discover prognostic multigene signatures from early breast cancer data. The partial Cox model combined with the multiple performance measures were used in both guiding the selection of the optimal panel of prognostic genes and prediction of risk within cross validation without dichotomising the follow-up times at any stage. The signatures were successfully externally cross validated in independent breast cancer datasets, yielding a hazard ratio of 2.55 [1.44, 4.51] for the top ranking signature.

Morphology and compression behaviour of biodegradable scaffolds produced by the sintering process

Porous poly(L-lactic acid) (PLA) scaffolds of 85 per cent and 90 per cent porosity are prepared using polymer sintering and porogen leaching method. Different weight fractions of 10 per cent, 30 per cent, and 50 per cent of hydroxyapatite (HA) are added to the PLA to control the acidity and degradation rate. The three-dimensional (3D) morphology and surface porosity are tested using micro-computer tomography (micro-CT), optical microscopy, and scanning electron microscopy (SEM). Results indicate that the surface porosity does not change on the addition of HA. The micro-CT examinations show a slight decrease in the pore size and increase in the wall thickness accompanied by reduced anisotropy for the scaffolds containing HA. Scanning electron micrographs show detectable interconnected pores for the scaffold with pure PLA. Addition of the HA results in agglomeration of the HA particles and reduced leaching of the porogen. Compression tests of the scaffold identify three stages in the stress-strain curve. The addition of HA results in a reduction in the modulus of the scaffold at the first stage of elastic bending of the wall, but this is reversed for the second and third stages of collapse of the wall and densification in the compression tests. In the scaffolds with 85 per cent porosity, the addition of a high percentage of HA could result in 70 per cent decrease in stiffness in the first stage, 200 per cent increase in stiffness in the second stage, and 20 per cent increase in stiffness in the third stage. The results of these tests are compared with the Gibson cellular material model that is proposed for prediction of the behaviour of cellular material under compression. The pH and molecular weight changes are tracked for the scaffolds within a period of 35 days. The addition of HA keeps the pH in the alkaline region, which results in higher rate of degradation at an early period of observation, followed by a reduced rate of degradation later in the process. The final molecular weight is higher for the scaffolds with HA than for scaffolds of pure PLA. The manufactured scaffolds offer acceptable properties in terms of the pore size range and interconnectivity of the pores and porosity for non-load-bearing bone graft substitute; however, improvement to the mixing of the phases of PLA and HA is required to achieve better integrity of the composite scaffolds. © 2008 IMechE.

A multi-output two-stage locally regularized model construction method using the extreme learning machine

This paper investigates the construction of linear-in-the-parameters (LITP) models for multi-output regression problems. Most existing stepwise forward algorithms choose the regressor terms one by one, each time maximizing the model error reduction ratio. The drawback is that such procedures cannot guarantee a sparse model, especially under highly noisy learning conditions. The main objective of this paper is to improve the sparsity and generalization capability of a model for multi-output regression problems, while reducing the computational complexity. This is achieved by proposing a novel multi-output two-stage locally regularized model construction (MTLRMC) method using the extreme learning machine (ELM). In this new algorithm, the nonlinear parameters in each term, such as the width of the Gaussian function and the power of a polynomial term, are firstly determined by the ELM. An initial multi-output LITP model is then generated according to the termination criteria in the first stage. The significance of each selected regressor is checked and the insignificant ones are replaced at the second stage. The proposed method can produce an optimized compact model by using the regularized parameters. Further, to reduce the computational complexity, a proper regression context is used to allow fast implementation of the proposed method. Simulation results confirm the effectiveness of the proposed technique. © 2013 Elsevier B.V.

Kinetostatic-model-based Stiffness Analysis of Exechon PKM

As a comparative newly-invented PKM with over-constraints in kinematic chains, the Exechon has attracted extensive attention from the research society. Different from the well-recognized kinematics analysis, the research on the stiffness characteristics of the Exechon still remains as a challenge due to the structural complexity. In order to achieve a thorough understanding of the stiffness characteristics of the Exechon PKM, this paper proposed an analytical kinetostatic model by using the substructure synthesis technique. The whole PKM system is decomposed into a moving platform subsystem, three limb subsystems and a fixed base subsystem, which are connected to each other sequentially through corresponding joints. Each limb body is modeled as a spatial beam with a uniform cross-section constrained by two sets of lumped springs. The equilibrium equation of each individual limb assemblage is derived through finite element formulation and combined with that of the moving platform derived with Newtonian method to construct the governing kinetostatic equations of the system after introducing the deformation compatibility conditions between the moving platform and the limbs. By extracting the 6 x 6 block matrix from the inversion of the governing compliance matrix, the stiffness of the moving platform is formulated. The computation for the stiffness of the Exechon PKM at a typical configuration as well as throughout the workspace is carried out in a quick manner with a piece-by-piece partition algorithm. The numerical simulations reveal a strong position-dependency of the PKM's stiffness in that it is symmetric relative to a work plane due to structural features. At the last stage, the effects of some design variables such as structural, dimensional and stiffness parameters on system rigidity are investigated with the purpose of providing useful information for the structural optimization and performance enhancement of the Exechon PKM. It is worthy mentioning that the proposed methodology of stiffness modeling in this paper can also be applied to other overconstrained PKMs and can evaluate the global rigidity over workplace efficiently with minor revisions.