Compact Model Representation Research Articles

Data-driven System Identification of Thermal Systems using Machine Learning

Abstract The paper addresses the identification of spatial-temporal mirror surface deformations as a result of laser-based heat load within the lithography process of integrated circuit production. The thermal diffusion and surface deformation are modeled by separation of the spatial-temporal effects using data-driven orthogonal decomposition. A novel tree adjoining grammar (TAG) and sparsity enhanced symbolic-regression-based learning methods are deployed to discover temporal dynamics that connect the spatial variation. The resulting data-driven procedure is applied to automatically synthetise a compact model representation of synthetic thermal effects induced mirror surface deformations.

Image Segmentation for Skin Detection

Human skin detection, which usually performed before image processing, is the method of discovering skin-colored pixels and regions that may be of human faces or limbs in videos or photos. Many computer vision approaches have been developed for skin detection. A skin detector usually transforms a given pixel into a suitable color space and then uses a skin classifier to mark the pixel as a skin or a non-skin pixel. A skin classifier explains the decision boundary of the class of a skin color in the color space based on skin-colored pixels. The purpose of this research is to build a skin detection system that will distinguish between skin and non-skin pixels in colored still pictures. This performed by introducing a metric that measures the distances of pixel colors to skin tones. Results showed that the YCbCr color space performed better skin pixel detection than regular Red Green Blue images due to its isolation of the overall energy of an image in the luminance band. The RGB color space poorly classified images with wooden backgrounds or objects. Then, a histogram-based image segmentation scheme utilized to distinguish between the skin and non-skin pixels. The need for a compact skin model representation should stimulate the development of parametric models of skin detection, which is a future research direction.

Shape, Pose and Density Statistical Model for 3D Reconstruction of Articulated Structures from X-Ray Images.

This article proposes a joint statistical model, to describe the volumetric shape + pose + density information, and a reconstruction algorithm to simultaneously recover the volumetric information of several anatomical structures from biplanar radiographs. A PCA-based representation is proposed as compact model representation and a hybrid AAM search and genetic optimization is used to perform the reconstruction. A study was conducted to recover a 3D volume grid containing a human knee mesh from 2 orthogonal simulated radiographs. The model was computed on a data set of 200 subjects and the reconstruction test was performed on 18 subjects, leading to a surface distance RMSE of 0.7 ± 0.31 mm for the distal femur, 0.9 ± 0.3 mm for the proximal tibia and 0.8 ± 0.3 mm for the fibula. These results demonstrate the feasibility and the pertinence of the proposed approach, the next step being its application in a clinical context.

A multi-port thermal coupling model for multi-chip power modules suitable for circuit simulators

This paper investigates two compact thermal model representations for multi-chip power modules, namely the thermal impedance matrix model and the thermal admittance matrix model. The latter can shape a multi-port thermal network without controlled temperature sources, and can be readily implemented in circuit simulators. The mutual transformation between the two models and their relationship to parameters in the multi-port network are revealed. In addition, practical tips for thermal model parameter extractions based on temperature measurements and curve-fitting are discussed. The multi-port thermal model is verified by simulations and experimental results. It confirms that more accurate temperature estimation can be achieved compared with the thermal model without the thermal coupling effect.

Statistical modeling and removal of lipid membrane projections for cryo-EM structure determination of reconstituted membrane proteins.

This paper describes steps in the single-particle cryo-EM 3D structure determination of membrane proteins in their membrane environment. Using images of the Kv1.2 potassium-channel complex reconstituted into lipid vesicles, we describe procedures for the merging of focal-pairs of exposures and the removal of the vesicle-membrane signal from the micrographs. These steps allow 3D reconstruction to be performed from the protein particle images. We construct a 2D statistical model of the vesicle structure based on higher-order singular value decomposition (HOSVD), by taking into account the structural symmetries of the vesicles in polar coordinates. Non-roundness in the vesicle structure is handled with a non-linear shape alignment to a reference, which ensures a compact model representation. The results show that the learned model is an accurate representation of the imaged vesicle structures. Precise removal of the strong membrane signals allows better alignment and classification of images of small membrane-protein particles, and allows higher-resolution 3D reconstruction.

Nonlinear Dynamic System Identification Using Volterra Series: Multi-Objective Optimization Approach

In this paper, identification of the nonlinear dynamic systems based on an optimized Volterra model structure is presented. Model structure selection is an important step in system identification, which involves the selection of variables and terms of a model. The key task is choosing a compact model representation where only significant terms are selected from among all the possible ones while also taking good performance into account. An automated algorithm based on multi-objective optimization is proposed for this purpose. The developed model should fulfill two criteria or objectives, namely, high prediction accuracy and optimum model structure. A genetic algorithm is applied to search the significant Volterra kernels from among all possible candidate model combinations. The result shows that the proposed algorithm is able to correctly identify the simulated examples and adequately model the nonlinear discrete dynamic system.

Web Recommendation Based on Unified Collaborative Filtering

Web usage mining technique is widely used for Web recommendation, which customizes Web content to user-preferred style. Traditional techniques of Web usage mining can only discover usage pattern explicitly. In order to employ the users’ feature and web pages’ attributes to get more accuracy recommendation, we propose a unified collaborative filtering model for web recommendation which combined the latent and external features of users and web page through back propagation neural networks. In the algorithm, we employ Probabilistic Latent Semantic Analysis (PLSA) method to get latent features. The main advantages of this technique over standard memory-based methods are the higher accuracy, constant time prediction, and an explicit and compact model representation. The preliminary experimental evaluation shows that substantial improvements in accuracy over existing methods can be obtained.

Unified collaborative filtering model based on combination of latent features

Collaborative filtering (CF) has been studied extensively in the literature and is demonstrated successfully in many different types of personalized recommender systems. In this paper, we propose a unified method combining the latent and external features of users and items for accurate recommendation. A mapping scheme for collaborative filtering problem to text analysis problem is introduced, and the probabilistic latent semantic analysis was used to calculate the latent features based on the historical rating data. The main advantages of this technique over standard memory-based methods are the higher accuracy, constant time prediction, and an explicit and compact model representation. The experimental evaluation shows that substantial improvements in accuracy over existing methods can be obtained.

Hybrid algorithm for compact and stable macromodelling of parameterised frequency responses

A novel hybrid technique for the macromodelling of parameterised frequency responses is presented. A more compact model representation is proposed that guarantees stability of the poles by construction. It is shown that accurate parametric macromodels can be calculated using a reduced amount of model coefficients and data samples. The obtained macromodels can be used for efficient real-time design space exploration, design optimisation and sensitivity analysis.

Probabilistic symbolic model checking with PRISM: a hybrid approach

In this paper we present efficient symbolic techniques for probabilistic model checking. These have been implemented in PRISM, a tool for the analysis of probabilistic models such as discrete-time Markov chains, continuous-time Markov chains and Markov decision processes using specifications in the probabilistic temporal logics PCTL and CSL. Motivated by the success of model checkers such as SMV which use BDDs (binary decision diagrams), we have developed an implementation of PCTL and CSL model checking based on MTBDDs (multi-terminal BDDs) and BDDs. Existing work in this direction has been hindered by the generally poor performance of MTBDD-based numerical computation, which is often substantially slower than explicit methods using sparse matrices. The focus of this paper is a novel hybrid technique which combines aspects of symbolic and explicit approaches to overcome these performance problems. For typical examples, we achieve a dramatic improvement over the purely symbolic approach. In addition, thanks to the compact model representation using MTBDDs, we can verify systems an order of magnitude larger than with sparse matrices, while almost matching or even beating them for speed.

Latent semantic models for collaborative filtering

Collaborative filtering aims at learning predictive models of user preferences, interests or behavior from community data, that is, a database of available user preferences. In this article, we describe a new family of model-based algorithms designed for this task. These algorithms rely on a statistical modelling technique that introduces latent class variables in a mixture model setting to discover user communities and prototypical interest profiles. We investigate several variations to deal with discrete and continuous response variables as well as with different objective functions. The main advantages of this technique over standard memory-based methods are higher accuracy, constant time prediction, and an explicit and compact model representation. The latter can also be used to mine for user communitites. The experimental evaluation shows that substantial improvements in accucracy over existing methods and published results can be obtained.

Extraction of a Dynamic Multiport Compact Thermal Model for a Silicon Microthruster

In this work, a methodology to extract dynamic multiport compact thermal model from finite element model (FEM) thermal simulations is presented. The proposed methodology is applied to an innovative silicon microthruster. The boundary conditions of the FEM model have been estimated from experimental measurements. The analysis of these transients with the methodology presented delivers a dynamic multiport compact RC model representation of the complete structure. The topology and sizing of the components is provided by a hybrid combination of Sequential Search algorithms together with nonlinear least squares optimization.

Modelling and estimating the pose of a human arm

Pose estimation of 3-D objects based on monocular computer vision is an ill-posed problem. To ease matters a model-based approach can be applied. Such an approach usually relies on iterating when matching the model and the image data. In this paper we estimate the 3-D pose of a human arm from a monocular image. To avoid the inherent problems when iterating, we apply an exhaustive matching strategy. To make this plausible, we reduce the size of the solution space through a very compact model representation of the arm and prune the solution space. The model is developed through a detailed investigation of the functionality and structure of the arm and the shoulder complex. The model consists of just two parameters and is based on the screw-axis representation together with image measurements. The pruning is achieved through kinematic constraints and it turns out that the solution space of the compact model can be pruned 97%, on average. Altogether, the compact representation and the constraints reduce the solution space significantly and, therefore, allow for an exhaustive matching procedure. The approach is tested in a model-based silhouette framework, and tests show promising results.

Thermal compact modeling of parallel plate heat sinks

Growing complexity of electronic systems has resulted in an increased computational effort in CFD modeling of electronic systems. To reduce the computational effort, one or several heat sinks can be represented by a compact "porous block" model, with an effective thermal conductivity and pressure loss coefficient. In this study of parallel plate heat sinks in laminar forced convection, a methodology is developed to rigorously determine the thermal properties of compact heat sink models that provide a high level of accuracy. The results of an extensive set of CFD simulations for a three heat sink channel covering two distinct heat sink geometries, air velocities from 0.25 m/s to 2 m/s and various spacings between the heat sinks, were used to create and evaluate the fidelity of compact models. The results of this study establish the validity and value in using the porous block compact model representation for noncritical heat sinks in an electronic assembly. The results also reveal that a location-independent porous-block representation can yield excellent agreement in the prediction of the thermal characteristics of state-of-the-art heat sinks.

Equivalent circuit modeling of static substrate thermal coupling using VCVS representation

A new method is described which allows substrate thermal coupling between active devices to be accurately represented in a circuit simulation environment. The method, based on a substrate thermal equivalent circuit containing resistors and voltage-controlled voltage sources, allows for exact representation of substrate thermal coupling at any number of evaluation points. The topology of the equivalent circuit and derivation of its coefficients is described, and application of the technique to inter- and intradevice thermal effects is illustrated. The method is applied with a simple self-heating compact model representation to a measured GaAs device characteristic exhibiting gain collapse, and is found to accurately predict electrothermal behavior.

Predictive worst case statistical modeling of 0.8- mu m BICMOS bipolar transistors: a methodology based on process and mixed device/circuit level simulators

The authors discuss the use of mixed-level physics-based device/circuit simulation software and semiconductor process simulator in the construction of predictive worst case process conditions for bipolar transistors currently being manufactured in AT&T 0.8- mu m BICMOS technology. Process fluctuations are introduced into the process simulator using the Latin hypercube (Monte Carlo) sampling method. The method is different from those in previous similar studies in that the compact device model parameter extraction step for each sample process is bypassed and active devices in the circuit are described by the physical device simulator rather than a compact model representation. This eliminates deficiencies associated with compact semiconductor device models. Furthermore, inaccuracies and difficulties introduced by compact model parameter extractions (especially for bipolar transistors) are also eliminated. The method is very useful in identifying critical process steps which determine the electrical performance of the devices and circuits.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>