A NEW DERIVATIVE FREE OPTIMIZATION METHOD BASED ON CONIC INTERPOLATION MODEL

In this paper, a new derivative free trust region method is developed based on the conic interpolation model for the unconstrained optimization. The conic interpolation model is built by means of the quadratic model function, the collinear scaling formula, quadratic approximation and interpolation. All the parameters in this model are determined by objective function interpolation condition. A new derivative free method is developed based upon this model and the global convergence of this new method is proved without any information on gradient.

Speaker Adaptation with Transformation Matrix Linear Interpolation

A transformation matrix linear interpolation (TMLI) approach for speaker adaptation is proposed. TMLI uses the transformation matrixes produced by MLLR from selected training speakers and the testing speaker. With only 3 adaptation sentences, the performance shows a 12.12% word error rate reduction. As the number of adaptation sentences increases, the performance saturates quickly. To improve the behavior of TMLI for large amounts of adaptation data, the TMLI+MAP method which combines TMLI with MAP technique is proposed. Experimental results show TMLI+MAP achieved better recognition accuracy than MAP and MLLR+MAP for both small and large amounts of adaptation data. Key words speech recognition - speaker adaptation - MLLR - MAP - maximum likelihood model interpolation (MLMI) CLC number TN 912. 34 Foundation item: Supported by the Science and Technology Committee of Shanghai (01JC14033)Biography: XU Xiang-hua (1977-), female, Ph. D. candidate, research direction: large vocabulary continuous Mandarin speech recognition and speaker adaptation

Implicit Shape Reconstruction of Unorganized Points Using PDE-Based Deformable 3D Manifolds

In this work we consider the problem of shape reconstruction from an unorganized data set which has many important applications in medical imaging, scientific computing, reverse engineering and geometric modelling. The reconstructed surface is obtained by continuously deforming an initial surface following the Partial Differential Equation (PDE)-based diffusion model derived by a minimal volume-like variational formulation. The evolution is driven both by the distance from the data set and by the curvature analytically computed by it. The distance function is computed by implicit local interpolants defined in terms of radial basis functions. Space discretization of the PDE model is obtained by finite co-volume schemes and semi-implicit approach is used in time/scale. The use of a level set method for the numerical computation of the surface reconstruction allows us to handle complex geometry and even changing topology,without the need of user-interaction. Numerical examples demonstrate the ability of the proposed method to produce high quality reconstructions. Moreover, we show the effectiveness of the new approach to solve hole filling problems and Boolean operations between different data sets.

A Combined Speaker Adaptation Method for Mandarin Speech Recognition

A speaker adaptation method that combines transformation matrix linear interpolation with maximum a posteriori (MAP) was proposed. Firstly this method can keep the asymptotical characteristic of MAP. Secondly, as the method uses linear interpolation with several speaker-dependent (SD) transformation matrixes, it can fully use the prior knowledge and keep fast adaptation. The experimental results show that the combined method achieves an 8.24% word error rate reduction with only one adaptation utterance, and keeps asymptotic to the performance of SD model for large amounts of adaptation data.

A New Direct Search Method Based on Separable Fractional Interpolation Model

In this paper, we propose a new separable fractional interpolation model which can be established by 2n interpolation points where n is the number of variables. Based on this model, a new direct search method is presented. In this method, a new iterate is determined by solving the fractional interpolation model in trust region. Under mild assumptions, the convergence results of this method are given and proved, Numerical experiments show that the new method is promising.

Heat Health Risk and Adaptability Assessments at the Subdistrict Scale in Metropolitan Beijing

Against the background of global climate change, the increasing heat health risk from the combined effect of changes in high temperature, exposure, vulnerability, and other factors has become a growing concern.Yet the low number of temperature observation stations is insufficient to represent the complex changes in urban heatwaves, and subdistrict-scale(town, township, neighborhood committee, and equivalent) heat health risk and adaptability assessments are still limited. In this study, we built daytime and nighttime high-temperature interpolation models supported by data from 225 meteorological stations in Beijing.The models performed well at interpolating the cumulative hours of high temperature and the interpolation quality at night was better than that during the day. We further established a methodological framework for heat health risk and adaptability assessments based on heat hazard, population exposure, social vulnerability, and adaptability at the subdistrict scale in Beijing. Our results show that the heat health risk hotspots were mainly located in the central urban area,with 81 hotspots during the day and 76 at night. The average value of the heat health risk index of urban areas was 5.60 times higher than that of suburban areas in the daytime,and 6.70 times higher than that of suburban areas in the night. Greater population density and higher intensity of heat hazards were the main reasons for the high risk in most heat health risk hotspots. Combined with a heat-adaptive-capacity evaluation for hotspot areas, this study suggests that 11high-risk and low-adaptation subdistricts are priority areas for government action to reduce heat health risk in policy formulation and urban development.

A multiobjective discrete combination optimization method for dynamics design of engineering structures

This paper presents a new multiobjective discrete optimization method for the engineering design of dynamic problems.A discrete combinatorial optimization problem is solved using a particle swarm optimization algorithm coupled with a stair‐form interpolation model.To address multiobjective optimization issues,a weighted average approach is implemented to convert the multiobjective optimization problem into an equivalent single‐objective optimization problem.Design con-straints are taken into consideration by using the penalty function strategy.The proposed method is first verified with a 10‐bar truss structure design problem,where the cross‐sectional area of each bar is optimized to minimize both volume and node displacement.Second,the dynamic issue for hybrid composite laminates is investigated by maximizing the fundamental frequency and minimizing the cost.The results reveal that the optimized results generated by the proposed method agree well with those from other approaches.

Numerical study of the flow in the Yellow River with non-monotonous banks

A 2-D depth averaged RNG k- ε model is developed to simulate the flow in a typical reach of the Upper Yellow River with non-monotonic banks. In order to take account of the effect of the secondary flow in a bend, the momentum equations are modified by adding an additional source term. A comparison between the numerical simulation and the field measurements indicates that the improved 2-D depth averaged RNG k- ε model can improve the accuracy of the numerical simulation. An arc spline interpolation method is developed to interpolate the non-monotonic river banks. The method can also be reasonably applied for the 2-D interpolation of the river bed level. Through a comparison of the water surface gradients simulated in the seven bends of the studied reach, some analytical formulae are improved to reasonably calculate the longitudinal and transverse gradients in meandering river reaches. Furthermore, the positions of the maximum water depth and the maximum velocity in a typical bend are discussed.