Drop In Accuracy Research Articles

High-order possibilistic c-means algorithms based on tensor decompositions for big data in IoT

Internet of Things (IoT) connects the physical world and the cyber world to offer intelligent services by data mining for big data. Each big data sample typically involves a large number of attributes, posing a remarkable challenge on the high-order possibilistic c-means algorithm (HOPCM). Specially, HOPCM requires high-performance servers with a large-scale memory and a powerful computing unit, to cluster big samples, limiting its applicability in IoT systems with low-end devices such as portable computing units and embedded devises which have only limited memory space and computing power. In this paper, we propose two high-order possibilistic c-means algorithms based on the canonical polyadic decomposition (CP-HOPCM) and the tensor-train network (TT-HOPCM) for clustering big data. In detail, we use the canonical polyadic decomposition and the tensor-train network to compress the attributes of each big data sample. To evaluate the performance of our algorithms, we conduct the experiments on two representative big data datasets, i.e., NUS-WIDE-14 and SNAE2, by comparison with the conventional high-order possibilistic c-means algorithm in terms of attributes reduction, execution time, memory usage and clustering accuracy. Results imply that CP-HOPCM and TT-HOPCM are potential for big data clustering in IoT systems with low-end devices since they can achieve a high compression rate for heterogeneous samples to save the memory space significantly without a significant clustering accuracy drop.

Balancing accuracy and diversity in recommendations using matrix completion framework

Design of recommender systems aimed at achieving high prediction accuracy is a widely researched area. However, several studies have suggested the need for diversified recommendations, with acceptable level of accuracy, to avoid monotony and improve customers’ experience. However, increasing diversity comes with an associated reduction in recommendation accuracy; thereby necessitating an optimum tradeoff between the two. In this work, we attempt to achieve accuracy-diversity balance, by exploiting available ratings and item metadata, through a single (joint) optimization model built over the matrix completion framework. Most existing works, unlike our formulation, propose a 2-stage model - a heuristic item ranking scheme on top of an existing collaborative filtering technique. Experimental evaluation on a movie recommender system indicates that our model achieves higher diversity for a given drop in accuracy as compared to existing state of the art techniques.

Proposal and analysis of different methodologies for the shading and blocking efficiency in central receivers systems

Nowadays, there are many simulation codes aimed at the simulation of the solar field performance and the optimization of the layout of central receivers systems. These codes should obtain fast and accurately the different efficiency factors of the solar field at different sun positions (representative of the yearly operation). Among these factors, the shading and blocking efficiency is maybe the most demanding one regarding the computational effort.In this paper four non-conventional methodologies are presented for the calculation of the shading and blocking efficiency. The codes have been developed with the ambition of decreasing the computational time without a significant accuracy drop. For that reason, they are suitable for optimization tools. Additionally, a new methodology for the determination of the subset of heliostats with potential for shading or blocking is presented.The performance of the methodologies is evaluated by means of a study of the errors and computational times, which are compared to those reached by a conventional Monte-Carlo ray tracing reference simulation. Results indicate that the proposed methodologies, particularly two of them, present good accuracy and a significant decrease of the computational time. The causes of the main errors of each methodology are also discussed.

Effect of Reduced Tube Voltage on Diagnostic Accuracy of CT Colonography.

The normal tube voltage in computed tomography colonography (CTC) is 120 kV. Some reports indicate that the use of a low tube voltage (lower than 120 kV) technique plays a significant role in reduction of radiation dose. However, to determine whether a lower tube voltage can reduce radiation dose without compromising diagnostic accuracy, an evaluation of images that are obtained while maintaining the volume CT dose index (CTDIvol) is required. This study investigated the effect of reduced tube voltage in CTC, without modifying radiation dose (i.e. constant CTDIvol), on image quality. Evaluation of image quality involved the shape of the noise power spectrum, surface profiling with volume rendering (VR), and receiver operating characteristic (ROC) analysis. The shape of the noise power spectrum obtained with a tube voltage of 80 kV and 100 kV was not similar to the one produced with a tube voltage of 120 kV. Moreover, a higher standard deviation was observed on volume-rendered images that were generated using the reduced tube voltages. In addition, ROC analysis revealed a statistically significant drop in diagnostic accuracy with reduced tube voltage, revealing that the modification of tube voltage affects volume-rendered images. The results of this study suggest that reduction of tube voltage in CTC, so as to reduce radiation dose, affects image quality and diagnostic accuracy.

Dempster-Shafer evidence theory for multi-bearing faults diagnosis

Support vector machines (SVMs) are frequently used in automated machinery faults diagnosis to classify multiple machinery faults by handling a high number of input features with low sampling data sets. SVMs are well known for fault detection that involves binary fault classifications only (i.e., healthy vs. faulty). However, when SVMs are used for multi-faults diagnostics and classification, they result in a drop in classification accuracy; this is because the adaptation of SVMs for multi-faults classifications requires the reduction of the multiple classification problem into multiple subsets of binary classification problems that result in many contradictory results from each individual SVM model. To overcome this problem, a novel SVM-DS (Dempster-Shafer evidence theory) model is proposed to resolve conflicting results generated from each SVM model and thus increase the classification accuracy. The analysis of results shows that the proposed SVM-DS model increased the accuracy of the fault diagnosis model from 76% to 94%, as SVM-DS continuously refines and eliminates all conflicting results from the original SVM model. The proposed SVM-DS model is found to be more accurate and effective in handling multi-faults diagnostic and classification problems commonly faced in the industries, as compared to the original SVM method.

StochasticNet in StochasticNet

Deep neural networks have been shown to outperform conventionalstate-of-the-art approaches in several structured predictionapplications. While high-performance computing devices such asGPUs has made developing very powerful deep neural networkspossible, it is not feasible to run these networks on low-cost, lowpowercomputing devices such as embedded CPUs or even embeddedGPUs. As such, there has been a lot of recent interestto produce efficient deep neural network architectures that can berun on small computing devices. Motivated by this, the idea ofStochasticNets was introduced, where deep neural networks areformed by leveraging random graph theory. It has been shownthat StochasticNet can form new networks with 2X or 3X architecturalefficiency while maintaining modeling accuracy. Motivated bythese promising results, here we investigate the idea of Stochastic-Net in StochasticNet (SiS), where highly-efficient deep neural networkswith Network in Network (NiN) architectures are formed ina stochastic manner. Such networks have an intertwining structurecomposed of convolutional layers and micro neural networksto boost the modeling accuracy. The experimental results showthat SiS can form deep neural networks with NiN architectures thathave 4X greater architectural efficiency with only a 2% dropin accuracy for the CIFAR10 dataset. The results are even morepromising for the SVHN dataset, where SiS formed deep neuralnetworks with NiN architectures that have 11.5X greater architecturalefficiency with only a 1% decrease in modeling accuracy.

Detecting Facial Retouching Using Supervised Deep Learning

Digitally altering, or retouching, face images is a common practice for images on social media, photo sharing websites, and even identification cards when the standards are not strictly enforced. This research demonstrates the effect of digital alterations on the performance of automatic face recognition, and also introduces an algorithm to classify face images as original or retouched with high accuracy. We first introduce two face image databases with unaltered and retouched images. Face recognition experiments performed on these databases show that when a retouched image is matched with its original image or an unaltered gallery image, the identification performance is considerably degraded, with a drop in matching accuracy of up to 25%. However, when images are retouched with the same style, the matching accuracy can be misleadingly high in comparison with matching original images. To detect retouching in face images, a novel supervised deep Boltzmann machine algorithm is proposed. It uses facial parts to learn discriminative features to classify face images as original or retouched. The proposed approach for classifying images as original or retouched yields an accuracy of over 87% on the data sets introduced in this paper and over 99% on three other makeup data sets used by previous researchers. This is a substantial increase in accuracy over the previous state-of-the-art algorithm, which has shown <50% accuracy in classifying original and retouched images from the ND-IIITD retouched faces database.

Decision analytic model of the diagnostic pathways for patients with suspected non-alcoholic fatty liver disease using non-invasive transient elastography and multiparametric magnetic resonance imaging

ObjectivesThe mortality associated with liver disease continues to increase, despite the improvements implemented in the UK healthcare as does the prevalence of non-alcoholic fatty liver disease (NAFLD), given the escalating...

Finding one face in a million [News

Helen of Troy may have had the face that launched a thousand ships, but even the best facial recognition algorithms might have had trouble finding her in a crowd of a million strangers. The first public benchmark test based on 1 million faces has shown how facial recognition algorithms from Google and other research groups around the world still fall well short of perfection. Facial recognition algorithms that had previously performed with more than 95 percent accuracy on a popular benchmark test involving 13,000 faces saw significant drops in accuracy when taking on the new MegaFace Challenge. The best performer, Google's FaceNet algorithm, dropped from near-perfect accuracy on the five-figure data set to 75 percent on the million-face test. Other top algorithms dropped from above 90 percent to below 60 percent. Some algorithms made the proper identification as seldom as 35 percent of the time.

Initialization of dynamic time warping using tree-based fast Nearest Neighbor

An efficient way to perform Dynamic Time Warping (DTW) search is by using the LBKeogh lower bound, which can eliminate a large number of candidate vectors out of the search process. Although effective, LBKeogh begins the DTW search using the first candidate vector, which is typically arbitrarily chosen. In this work, we propose initializing the LBKeogh-based DTW search using the Euclidean Distance Nearest Neighbor, derived by a fast tree-based Nearest Neighbor technique. Our experimental results suggest that, on one hand, this simple NN-based approach is quite accurate for trajectory classification of digit and letter gesturing and can initialize the DTW search very efficiently, thus requiring about 20% less search time than existing DTW implementations without any drop in recognition accuracy.

Skeleton-based canonical forms for non-rigid 3D shape retrieval

The retrieval of non-rigid 3D shapes is an important task. A common technique is to simplify this problem to a rigid shape retrieval task by producing a bending invariant canonical form for each shape in the dataset to be searched. It is common for these techniques to attempt to ``unbend'' a shape by applying multidimensional scaling to the distances between points on the mesh, but this leads to unwanted local shape distortions. We instead perform the unbending on the skeleton of the mesh, and use this to drive the deformation of the mesh itself. This leads to a computational speed-up and less distortions of the local details of the shape. We compare our method against other canonical forms and our experiments show that our method achieves state-of-the-art retrieval accuracy in a recent canonical forms benchmark, and only a small drop in retrieval accuracy over state-of-the-art in a second recent benchmark, while being significantly faster.

The BRaliBase dent-a tale of benchmark design and interpretation.

BRaliBase is a widely used benchmark for assessing the accuracy of RNA secondary structure alignment methods. In most case studies based on the BRaliBase benchmark, one can observe a puzzling drop in accuracy in the 40–60% sequence identity range, the so-called ‘BRaliBase Dent’. In this article, we show this dent is owing to a bias in the composition of the BRaliBase benchmark, namely the inclusion of a disproportionate number of transfer RNAs, which exhibit a conserved secondary structure. Our analysis, aside of its interest regarding the specific case of the BRaliBase benchmark, also raises important questions regarding the design and use of benchmarks in computational biology.

Shoot to Know What: An Application of Deep Networks on Mobile Devices

Convolutional neural networks (CNNs) have achieved impressive performance in a wide range of computer vision areas. However, the application on mobile devices remains intractable due to the high computation complexity. In this demo, we propose the Quantized CNN (Q-CNN), an efficient framework for CNN models, to fulfill efficient and accurate image classification on mobile devices. Our Q-CNN framework dramatically accelerates the computation and reduces the storage/memory consumption, so that mobile devices can independently run an ImageNet-scale CNN model. Experiments on the ILSVRC-12 dataset demonstrate 4~6x speed-up and 15~20x compression, with merely one percentage drop in the classification accuracy. Based on the Q-CNN framework, even mobile devices can accurately classify images within one second.

Accuracy feedback improves word learning from context: evidence from a meaning-generation task

The present study asked whether accuracy feedback on a meaning generation task would lead to improved contextual word learning (CWL). Active generation can facilitate learning by increasing task engagement and memory retrieval, which strengthens new word representations. However, forced generation results in increased errors, which can be detrimental for learning if they are not corrected. The goal of this study was to determine whether immediate feedback on response accuracy would ameliorate this risk. The study was conducted using an intelligent tutoring system, which presents target words in multiple contexts and prompts users to generate a target word meaning after each context. One group of participants (feedback group) received immediate feedback based on Markov Estimation of Semantic Association (MESA), which estimates the distance between the learner response and the target word meaning. The control group did not receive feedback. Results from conventional (pre/post-test) measures showed greater gains in accuracy and confidence for the feedback group. Moreover, when contextual support was decreased mid-way through the training (from trial 3 to trial 4), MESA measures showed a corresponding drop in accuracy, but only for the No-Feedback group. These findings suggest that accuracy feedback can improve outcomes in CWL, particularly when there is an increased risk of errors. This strengthens the case for meaning generation as a tool to build high-quality lexical representations.

Physiological and Movement Demands of Rugby League Referees: Influence on Penalty Accuracy.

Research into the physiological and movement demands of Rugby League (RL) referees is limited, with only 1 study in the European Super League (SL). To date, no studies have considered decision making in RL referees. The purpose of this study was to quantify penalty accuracy scores of RL referees and to determine the relationship between penalty accuracy and total distance covered (TD), high-intensity running (HIR), and heart rate per 10-minute period of match play. Time motion analysis was undertaken on 8 referees over 148 European SL games during the 2012 season using 10-Hz global positioning system analysis and heart rate monitors. The number and timing of penalties awarded was quantified using Opta Stats. Referees awarded the correct decision on 74 ± 5% of occasions. Lowest accuracy was observed in the last 10-minute period of the game (67 ± 13%), with a moderate drop (effect size = 0.86) in accuracy observed between 60-70 minutes and 70-80 minutes. Despite this, there were only small correlations observed between mean heart rate, TD, HIR efforts, and penalty accuracy. Although a moderate correlation was observed between maximum velocity and accuracy. Despite only small correlations observed, it would be rash to assume that physiological and movement demands of refereeing have no influence on decision making. More likely, other confounding variables influence referee decision-making accuracy, requiring further investigation. Findings can be used by referees and coaches to inform training protocols, ensuring training is specific to both cognitive and physical match demands.

Attentional performance, impulsivity, and related neurotransmitter systems in apoE2, apoE3, and apoE4 female transgenic mice.

The apolipoprotein E (apoE) genotype influences cognitive performance in humans depending on age and sex. While the detrimental role of the apoE4 isoform on spatial learning and memory has been well-established in humans and rodents, less is known on its impact on the executive functions. We aimed to evaluate the effect of apoE isoforms (apoE2, apoE3, apoE4) on visuospatial attention and inhibitory control performance in female transgenic mice, and to determine the neurochemical and neuropharmacological basis of this potential relationship. Female mice carrying apoE2, apoE3, and apoE4 were trained in the five-choice serial reaction time task (5-CSRTT). Upon a stable performance, we manipulated the inter-trial interval and the stimulus duration to elicit impulsive responding and engage attention respectively. We further performed a pharmacological challenge by administering cholinergic and GABAergic agents. Finally, we analyzed the levels of brain amino acids and monoamines by using reversed phase high-performance liquid chromatography (HPLC). ApoE4 mice showed a deficient inhibitory control as revealed by increased perseveration and premature responding. When attention was challenged, apoE4 mice also showed a higher drop in accuracy. The adverse effect of scopolamine on the task was attenuated in apoE4 mice compared to apoE2 and apoE3. Furthermore, apoE4 mice showed less dopamine in the frontal cortex than apoE2 mice. We confirmed that the apoE genotype influences attention and inhibitory control in female transgenic mice. The influence of apoE isoforms in the brain neuromodulatory system may explain the cognitive and behavioral differences attributable to the genotype.

Confidence and Eyewitness Identifications: The Cross‐Race Effect, Decision Time and Accuracy

SummaryParticipants encountered same‐race and cross‐race faces at encoding, completed a series of line‐up identification tests and provided confidence ratings by using one of nine different confidence scales. Confidence was less well calibrated with identification accuracy when participants selected a cross‐race than a same‐race face because of overconfidence. By contrast, there was no cross‐race effect on confidence–accuracy calibration when participants responded ‘not present’. Whereas confidence was a very strong predictor of accuracy for fast identifications of a line‐up face, this was much less the case for slower decisions. Highly confident identifications showed a dramatic drop in accuracy from faster decisions to slower decisions, whereas there was little change in accuracy between faster and slower decisions for moderately confident or weakly confident identifications. Finally, we observed little influence of the format of the nine different confidence scales: numerical and verbal scales produced comparable calibration scores, as did scales with few or many points. Copyright © 2015 John Wiley &amp; Sons, Ltd.

DECIPHER: harnessing local sequence context to improve protein multiple sequence alignment.

BackgroundAlignment of large and diverse sequence sets is a common task in biological investigations, yet there remains considerable room for improvement in alignment quality. Multiple sequence alignment programs tend to reach maximal accuracy when aligning only a few sequences, and then diminish steadily as more sequences are added. This drop in accuracy can be partly attributed to a build-up of error and ambiguity as more sequences are aligned. Most high-throughput sequence alignment algorithms do not use contextual information under the assumption that sites are independent. This study examines the extent to which local sequence context can be exploited to improve the quality of large multiple sequence alignments.ResultsTwo predictors based on local sequence context were assessed: (i) single sequence secondary structure predictions, and (ii) modulation of gap costs according to the surrounding residues. The results indicate that context-based predictors have appreciable information content that can be utilized to create more accurate alignments. Furthermore, local context becomes more informative as the number of sequences increases, enabling more accurate protein alignments of large empirical benchmarks. These discoveries became the basis for DECIPHER, a new context-aware program for sequence alignment, which outperformed other programs on large sequence sets.ConclusionsPredicting secondary structure based on local sequence context is an efficient means of breaking the independence assumption in alignment. Since secondary structure is more conserved than primary sequence, it can be leveraged to improve the alignment of distantly related proteins. Moreover, secondary structure predictions increase in accuracy as more sequences are used in the prediction. This enables the scalable generation of large sequence alignments that maintain high accuracy even on diverse sequence sets. The DECIPHER R package and source code are freely available for download at DECIPHER.cee.wisc.edu and from the Bioconductor repository.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-015-0749-z) contains supplementary material, which is available to authorized users.

Accurate Radial Vaneless Diffuser One-Dimensional Model

A simplified one-dimensional model for the performance estimation of vaneless radial diffusers is presented. The starting point of such a model is that angular momentum losses occurring in vaneless diffusers are usually neglected in the most common turbomachinery textbooks: It is assumed that the angular momentum is conserved inside a vaneless diffuser, although a nonisentropic pressure transformation is considered at the same time. This means that fluid-dynamic losses are taken into account only for what concerns pressure recovery, whereas the evaluation of the outlet tangential velocity incoherently follows an ideal behavior. Several attempts were presented in the past in order to consider the loss of angular momentum, mainly solving a full set of differential equations based on the various developments of the initial work by Stanitz (1952, “One-Dimensional Compressible Flow in Vaneless Diffusers of Radial or Mixed-Flow Centrifugal Compressors, Including Effects of Friction, Heat Transfer and Area Change,” Report No. NACA TN 2610). However, such formulations are significantly more complex and are based on two empirical or calibration coefficients (skin friction coefficient and dissipation or turbulent mixing loss coefficient) which need to be properly assessed. In the present paper, a 1D model for diffuser losses computation is derived considering a single loss coefficient, and without the need of solving a set of differential equations. The model has been validated against massive industrial experimental campaigns, in which several diffuser geometries and operating conditions have been considered. The obtained results confirm the reliability of the proposed approach, able to predict the diffuser performance with negligible drop of accuracy in comparison with more sophisticated techniques. Both preliminary industrial designs and experimental evaluations of the diffusers may benefit from the proposed model.

Honking is just noise (or just about): The effect of energetic masking on recognition memory for spoken words

Previous research indicates that listeners encode both linguistic and indexical specifications of the speech signal in memory. Recent evidence suggests that non-linguistic sounds co-occurring with spoken words are also incorporated in our lexical memory. We argue that this “sound-specificity effect” might not be due so much to a word-sound association as to the different acoustic glimpses of the words that the associated sounds create. In several recognition-memory experiments, we paired spoken words with one of two car honk sounds and varied the level of energetic masking from exposure to test. We did not observe a drop in recognition accuracy for previously heard words when the paired sound changed as long as energetic masking was controlled. However, when we manipulated the temporal overlap between words and honking to create an energetic masking contrast, accuracy dropped. The finding suggests that listeners encode irrelevant non-speech information in memory, but only in certain contexts. Calling for an expansion of the mental lexicon to include non-speech auditory information might be premature. Current work is investigating the effect in non-native listeners of English, and whether maskers that are more integral to the words and hence more difficult to segregate lead to a more robust effect.