Smaller Model Size Research Articles

Tradeoffs in modeling performance of highly configurable software systems

Modeling the performance of a highly configurable software system requires capturing the influences of its configuration options and their interactions on the system’s performance. Performance-influence models quantify these influences, explaining this way the performance behavior of a configurable system as a whole. To be useful in practice, a performance-influence model should have a low prediction error, small model size, and reasonable computation time. Because of the inherent tradeoffs among these properties, optimizing for one property may negatively influence the others. It is unclear, though, to what extent these tradeoffs manifest themselves in practice, that is, whether a large configuration space can be described accurately only with large models and significant resource investment. By means of 10 real-world highly configurable systems from different domains, we have systematically studied the tradeoffs between the three properties. Surprisingly, we found that the tradeoffs between prediction error and model size and between prediction error and computation time are rather marginal. That is, we can learn accurate and small models in reasonable time, so that one performance-influence model can fit different use cases, such as program comprehension and performance prediction. We further investigated the reasons for why the tradeoffs are marginal. We found that interactions among four or more configuration options have only a minor influence on the prediction error and that ignoring them when learning a performance-influence model can save a substantial amount of computation time, while keeping the model small without considerably increasing the prediction error. This is an important insight for new sampling and learning techniques as they can focus on specific regions of the configuration space and find a sweet spot between accuracy and effort. We further analyzed the causes for the configuration options and their interactions having the observed influences on the systems’ performance. We were able to identify several patterns across subject systems, such as dominant configuration options and data pipelines, that explain the influences of highly influential configuration options and interactions, and give further insights into the domain of highly configurable systems.

Deep learning-based Diabetic Retinopathy assessment on embedded system.

Diabetic Retinopathy (DR) is a disease which affect the vision ability. The observation by an ophthalmologist usually conducted by analyzing the retinal images of the patient which are marked by some DR features. However some misdiagnosis are usually found due to human error. Here, a deep learning-based low-cost embedded system is established to assist the doctor for grading the severity of the DR from the retinal images. A compact deep learning algorithm named Deep-DR-Net which fits on a small embedded board is afterwards proposed for such purposes. In the heart of Deep-DR-Net, a cascaded encoder-classifier network is arranged using residual style for ensuring the small model size. The usage of different types of convolutional layers subsequently guarantees the features richness of the network for differentiating the grade of the DR. Experimental results show the capability of the proposed system for detecting the existence as well as grading the severity of the DR symptomps.

An anatomically-constrained local deformation model for monocular face capture

We present a new anatomically-constrained local face model and fitting approach for tracking 3D faces from 2D motion data in very high quality. In contrast to traditional global face models, often built from a large set of blendshapes, we propose a local deformation model composed of many small subspaces spatially distributed over the face. Our local model offers far more flexibility and expressiveness than global blendshape models, even with a much smaller model size. This flexibility would typically come at the cost of reduced robustness, in particular during the under-constrained task of monocular reconstruction. However, a key contribution of this work is that we consider the face anatomy and introduce subspace skin thickness constraints into our model, which constrain the face to only valid expressions and helps counteract depth ambiguities in monocular tracking. Given our new model, we present a novel fitting optimization that allows 3D facial performance reconstruction from a single view at extremely high quality, far beyond previous fitting approaches. Our model is flexible, and can be applied also when only sparse motion data is available, for example with marker-based motion capture or even face posing from artistic sketches. Furthermore, by incorporating anatomical constraints we can automatically estimate the rigid motion of the skull, obtaining a rigid stabilization of the performance for free. We demonstrate our model and single-view fitting method on a number of examples, including, for the first time, extreme local skin deformation caused by external forces such as wind, captured from a single high-speed camera.

A Bi-Projection Neural Network for Solving Constrained Quadratic Optimization Problems.

In this paper, a bi-projection neural network for solving a class of constrained quadratic optimization problems is proposed. It is proved that the proposed neural network is globally stable in the sense of Lyapunov, and the output trajectory of the proposed neural network will converge globally to an optimal solution. Compared with existing projection neural networks (PNNs), the proposed neural network has a very small model size owing to its bi-projection structure. Furthermore, an application to data fusion shows that the proposed neural network is very effective. Numerical results demonstrate that the proposed neural network is much faster than the existing PNNs.

A Fast Adaptive Tunable RBF Network For Nonstationary Systems.

This paper describes a novel on-line learning approach for radial basis function (RBF) neural network. Based on an RBF network with individually tunable nodes and a fixed small model size, the weight vector is adjusted using the multi-innovation recursive least square algorithm on-line. When the residual error of the RBF network becomes large despite of the weight adaptation, an insignificant node with little contribution to the overall system is replaced by a new node. Structural parameters of the new node are optimized by proposed fast algorithms in order to significantly improve the modeling performance. The proposed scheme describes a novel, flexible, and fast way for on-line system identification problems. Simulation results show that the proposed approach can significantly outperform existing ones for nonstationary systems in particular.

Probabilistic Vibration Analysis of a Mistuned Integrally Bladed Disk

Abstract. The objective of this paper is to probabilistically evaluate the effects of mistuned sectors on the dynamic characteristics of an integrally bladed disk (blisk). Small blade to blade physical variation in a disk is termed as mistuning. In this study, the dynamic characteristics of the perfectly tuned blisk were firstly analysed as a baseline. Secondly, a probabilistic approach is used for the mistuning analysis of a blisk. A reduced-order method named as the subset of nominal modes (SNM) was used to generate modes for a mistuned blisk from a cyclic perfectly tuned FE model without creating the full model. Furthermore, as only the modes with natural frequencies close to the modes of interest were considered, a relatively shorter computational time and a much smaller model size than a full blisk model is used. Therefore, the dynamic characteristics of the forced response for random mistuned blisks were obtained.

HColonies: a new hybrid metaheuristic for medical data classification

Medical data feature a number of characteristics that make their classification a complex task. Yet, the societal significance of the subject and the computational challenge it presents has caused the classification of medical datasets to be a popular research area. A new hybrid metaheuristic is presented for the classification task of medical datasets. The hybrid ant–bee colonies (HColonies) consists of two phases: an ant colony optimization (ACO) phase and an artificial bee colony (ABC) phase. The food sources of ABC are initialized into decision lists, constructed during the ACO phase using different subsets of the training data. The task of the ABC is to optimize the obtained decision lists. New variants of the ABC operators are proposed to suit the classification task. Results on a number of benchmark, real-world medical datasets show the usefulness of the proposed approach. Classification models obtained feature good predictive accuracy and relatively small model size.

Boosted Mixture Learning of Gaussian Mixture Hidden Markov Models Based on Maximum Likelihood for Speech Recognition

In this paper, we apply the well-known boosted mixture learning (BML) method to learn Gaussian mixture HMMs in speech recognition. BML is an incremental method to learn mixture models for classification problems. In each step of BML, one new mixture component is estimated according to the functional gradient of an objective function to ensure that it is added along the direction that maximizes the objective function. Several techniques have been proposed to extend BML from simple mixture models like the Gaussian mixture model (GMM) to the Gaussian mixture hidden Markov model (HMM), including Viterbi approximation for state segmentation, weight decay and sampling boosting to initialize sample weights to avoid overfitting, combination between partial updating and global updating to refine model parameters in each BML iteration, and use of the Bayesian Information Criterion (BIC) for parsimonious modeling. Experimental results on two large-vocabulary continuous speech recognition tasks, namely the WSJ-5k and Switchboard tasks, have shown that the proposed BML yields significant performance gain over the conventional training procedure, especially for small model sizes.

Investment portfolios under uncertainty for utilizing natural gas resources

Numerous reasons including lower carbon and sulfur emissions have led to the rapid growth of natural gas (NG) demand. However, more than one-third of world NG reserves are stranded, i.e., either remote (e.g., offshore) or in regions with saturated markets. This reality makes the investment decisions complex and uncertain for NG field developers. In this study, we consider the case of a company that wishes to develop a stranded natural gas reserve for some potential nearby markets under uncertain prices of crude oil and feed gas, and demands of liquefied NG (LNG), compressed NG (CNG), and gas-to-liquid (GTL) products. We present a 2-stage stochastic mixed-integer linear program (MILP) that yields maximum-ENPV (expected NPV) decisions on production capacities, market allocations, and delivery vessels. The small model size allows us to consider many stochastic scenarios in our scenario-based approach. We illustrate our approach using several examples.

Content-based File-type Identification Using Cosine Similarity and a Divide-and-Conquer Approach

Identifying the file type (TXT, EXE, JPEG, etc.) is important for computer security applications such as computer forensics, steganalysis, and antivirus programs. The common approach for this is to use file extensions, magic numbers, or other header information. However, these are susceptible to tampering or corruption; for instance, the file extension can be easily spoofed and the magic numbers can be obfuscated. A more reliable approach may be to analyze the file content instead of using only the tip of the information (metadata). This paper proposes two methods based on the file content. First, we use the cosine distance as a similarity metric when comparing the file content rather than the Mahalanobis distance that is popular and has been used by the other related approaches. The cosine similarity (unlike the Mahalanobis distance) retains the classification accuracy on a small number of highly frequent byte patterns which leads to a smaller model size and faster detection rate. Second, we decompose the identification procedure into two steps by taking the divide and conquer: in the first step, the similar files in terms of byte pattern frequencies are grouped into several clusters. In the next step, the cluster which contains different file types is fed to the neural network in order for finer classification. The experiments showed that the classification followed by clustering leads to higher accuracies.

The Clustered Causal State Algorithm: Efficient Pattern Discovery for Lossy Data-Compression Applications

Pattern discovery is a potential boon for data compression. By identifying generic patterns without human supervision, pattern discovery algorithms can extract the most relevant information for greatest fidelity in lossy compression. However, current approaches to pattern discovery are inefficient and produce cumbersome descriptions of patterns. The Clustered Causal State Algorithm (CCSA) is a new pattern discovery algorithm incorporating recent clustering technology. This algorithm sacrifices accuracy for increased efficiency and smaller model sizes. This makes CCSA ideal for lossy data compression and other real-time applications. This algorithm is compared to other pattern discovery algorithms and demonstrated in an image compression application.

Boosting image classification with LDA-based feature combination for digital photograph management

Image classification is of great importance for digital photograph management. In this paper we propose a general statistical learning method based on boosting algorithm to perform image classification for photograph annotation and management. The proposed method employs both features extracted from image content (i.e., color moment and edge direction histogram) and features from the EXIF metadata recorded by digital cameras. To fully utilize potential feature correlations and improve the classification accuracy, feature combination is needed. We incorporate linear discriminant analysis (LDA) algorithm to implement linear combinations between selected features and generate new combined features. The combined features are used along with the original features in boosting algorithm for improving classification performance. To make the proposed learning algorithm more efficient, we present two heuristics for selective feature combinations, which can significantly reduce training computation without losing performance. The proposed image classification method has several advantages: small model size, computational efficiency and improved classification performance based on LDA feature combination.

방파제를 고려한 초대형 부유식 해상구조물의 유탄성 응답 특성에 관한 실험적 연구

In this paper, an experimental study on the hydroelastic behaviors of a VLFS with L=5,000m was made considering a breakwater. The principal dimensions of the VLFS model were 9m<TEX>${\times}$</TEX>1.8<TEX>${\times}$</TEX>0.0108m(L<TEX>${\times}$</TEX>B<TEX>${\times}$</TEX>D) and the length of breakwater was 12.6m (1.4L). The distance between the VLFS and the breakwater varied from B/2 to 28. The wide tank test results were compared with the numerical predictions and the comparison showed a little gap along its longitudinal axis, in spite of using the very small model size due to the scale 1/555.5

PIV application in advanced low reynolds number facility

The advent of micro aerial vehicles (MAV) and advanced flow control methods, using microelectromechanical systems (MEMS) and micro-blowing or suction, has significantly increased the need for a better understanding of very low Reynolds number flows. Aerodynamic testing in this flow regime using conventional facilities, such as wind or water tunnels, is seriously hampered by the necessary small model sizes and low testing speeds, rendering it very difficult to diagnose the flow with adequate spatial resolution and to measure aerodynamic loads with sufficient accuracy. A proof-of-concept pilot of a novel facility specifically intended for low Reynolds number testing has been built and is briefly described. It is based on the use of a high kinematic viscosity test medium, which allows testing superscale models at the correct Reynolds number. The facility is akin to a towing tank, but does not have a free surface with its associated shortcomings. Implementation of the particle image velocimetry (PIV) technique in this type of facility is complicated by the fact that the optics, rather than being stationary as is the case in conventional aerodynamic testing facilities, move in synchronism with the model. Mistracking directly results in velocity-field errors which must be corrected. Methods to do so are described.

Nodal Constraint, Shear Deformation and Continuity Effects Related to the Modeling of Debonding of Laminates, Using Plate Elements

The effects of several critical assumptions and parameters on the computation of strain energy release rates for delamination and debond configurations modeled with plate elements have been quantified. The method of calculation is based on the virtual crack closure technique (VCCT), and models of the upper and lower surface of the delamination or debond that use two-dimensional (2D) plate elements rather than three-dimensional (3D) solid elements. The major advantages of the plate element modeling technique are a smaller model size and simpler configurational modeling. Specific issues that are discussed include: constraint of translational degrees of freedom, rotational degrees of freedom or both in the neighborhood of the debond front, shear deformation assumptions; and continuity of material properties and section stiffness in the vicinity of the debond front. Where appropriate, the plate element analyses are compared with corresponding two-dimensional plane strain analyses.

Scalable feature selection, classification and signature generation for organizing large text databases into hierarchical topic taxonomies

We explore how to organize large text databases hierarchically by topic to aid better searching, browsing and filtering. Many corpora, such as internet directories, digital libraries, and patent databases are manually organized into topic hierarchies, also called taxonomies. Similar to indices for relational data, taxonomies make search and access more efficient. However, the exponential growth in the volume of on-line textual information makes it nearly impossible to maintain such taxonomic organization for large, fast-changing corpora by hand. We describe an automatic system that starts with a small sample of the corpus in which topics have been assigned by hand, and then updates the database with new documents as the corpus grows, assigning topics to these new documents with high speed and accuracy. To do this, we use techniques from statistical pattern recognition to efficiently separate the feature words, or discriminants, from thenoise words at each node of the taxonomy. Using these, we build a multilevel classifier. At each node, this classifier can ignore the large number of “noise” words in a document. Thus, the classifier has a small model size and is very fast. Owing to the use of context-sensitive features, the classifier is very accurate. As a by-product, we can compute for each document a set of terms that occur significantly more often in it than in the classes to which it belongs. We describe the design and implementation of our system, stressing how to exploit standard, efficient relational operations like sorts and joins. We report on experiences with the Reuters newswire benchmark, the US patent database, and web document samples from Yahoo!. We discuss applications where our system can improve searching and filtering capabilities.

Computational effects of inlet representation on powered hypersonic,airbreathing models

Computational results are presented to illustrate the powered aftbody effects of representing the scramjet inlet on a generic hypersonic vehicle with a fairing, to divert the external flow, as compared to an operating flow-through scramjet inlet. This study is pertinent to the ground testing of hypersonic, airbreathing models employing scramjet exhaust flow simulation in typical small-scale hypersonic wind tunnels. The comparison of aftbody effects due to inlet representation is well-suited for computational study, since small model size typically precludes the ability to ingest flow into the inlet and perform exhaust simulation at the same time. Two-dimensional analysis indicates that, although flowfield differences exist for the two types of inlet representations, little, if any, difference in surface aftbody characteristics is caused by fairing over the inlet.

Nosocomial toxic-shock syndrome in two patients after knee surgery

Redundant robots motion planning and control in uncertain task is addressed by model-based approach. Instead of control and adapt the environment to the robot or apply a complex visual servoing system, we have modeled the redundancy resolution (RR) on the parameter spaces that quantify uncertainties of the task. A modeling tool was Successive Approximations (SA). It provides very advantageous properties: small computational effort and small model size, accurate output, extrapolation, and generalization across parameter set obtained by random addressing of the model. The task discussed is press loading with typical two-dimensional uncertainties in pick-up and unloading locations. The robot used is 4 DOF planar robot. The SA-based models of redundancy resolution in the 2D parameter spaces are highly efficient: for more that 30 times less computational efforts resulted in a zero end-point errors, regardless of the task uncertainty.

Tapered Inlets for Pipe Culverts

The performance of pipe culverts on steep slopes with tapered inlets is, in general, sensitive to the presence of air-carrying vortices over the inlet. Under such conditions, an area of operation is defined. The minimum bound of the performance is shown to approximate inlet control operation of the inlet with atmospheric pressure over a free surface flow immediately downstream of the inlet control section. The performance of pipe culvert inlets is presented in terms of a discharge coefficient that closely approximates commonly used values of the discharge coefficient for orifices of similar edge geometry. The significance of various aspects of the inlet geometry with respect to the operation of the culvert, as well as to the location of the applicable control section, is explored. At small model sizes, the performance of pipe culverts with tapered inlets is shown to be subject to significant scale effects.