86 resultados para variance ratio
Resumo:
This paper proposes the use of Bayesian approaches with the cross likelihood ratio (CLR) as a criterion for speaker clustering within a speaker diarization system, using eigenvoice modeling techniques. The CLR has previously been shown to be an effective decision criterion for speaker clustering using Gaussian mixture models. Recently, eigenvoice modeling has become an increasingly popular technique, due to its ability to adequately represent a speaker based on sparse training data, as well as to provide an improved capture of differences in speaker characteristics. The integration of eigenvoice modeling into the CLR framework to capitalize on the advantage of both techniques has also been shown to be beneficial for the speaker clustering task. Building on that success, this paper proposes the use of Bayesian methods to compute the conditional probabilities in computing the CLR, thus effectively combining the eigenvoice-CLR framework with the advantages of a Bayesian approach to the diarization problem. Results obtained on the 2002 Rich Transcription (RT-02) Evaluation dataset show an improved clustering performance, resulting in a 33.5% relative improvement in the overall Diarization Error Rate (DER) compared to the baseline system.
Resumo:
One of the primary desired capabilities of any future air traffic separation management system is the ability to provide early conflict detection and resolution effectively and efficiently. In this paper, we consider the risk of conflict as a primary measurement to be used for early conflict detection. This paper focuses on developing a novel approach to assess the impact of different measurement uncertainty models on the estimated risk of conflict. The measurement uncertainty model can be used to represent different sensor accuracy and sensor choices. Our study demonstrates the value of modelling measurement uncertainty in the conflict risk estimation problem and presents techniques providing a means of assessing sensor requirements to achieve desired conflict detection performance.
Resumo:
The giant freshwater prawn (Macrobrachium rosenbergii) or GFP is one of the most important freshwater crustacean species in the inland aquaculture sector of many tropical and subtropical countries. Since the 1990’s, there has been rapid global expansion of freshwater prawn farming, especially in Asian countries, with an average annual rate of increase of 48% between 1999 and 2001 (New, 2005). In Vietnam, GFP is cultured in a variety of culture systems, typically in integrated or rotational rice-prawn culture (Phuong et al., 2006) and has become one of the most common farmed aquatic species in the country, due to its ability to grow rapidly and to attract high market price and high demand. Despite potential for expanded production, sustainability of freshwater prawn farming in the region is currently threatened by low production efficiency and vulnerability of farmed stocks to disease. Commercial large scale and small scale GFP farms in Vietnam have experienced relatively low stock productivity, large size and weight variation, a low proportion of edible meat (large head to body ratio), scarcity of good quality seed stock. The current situation highlights the need for a systematic stock improvement program for GFP in Vietnam aimed at improving economically important traits in this species. This study reports on the breeding program for fast growth employing combined (between and within) family selection in giant freshwater prawn in Vietnam. The base population was synthesized using a complete diallel cross including 9 crosses from two local stocks (DN and MK strains) and a third exotic stock (Malaysian strain - MY). In the next three selection generations, matings were conducted between genetically unrelated brood stock to produce full-sib and (paternal) half-sib families. All families were produced and reared separately until juveniles in each family were tagged as a batch using visible implant elastomer (VIE) at a body size of approximately 2 g. After tags were verified, 60 to 120 juveniles chosen randomly from each family were released into two common earthen ponds of 3,500 m2 pond for a grow-out period of 16 to 18 weeks. Selection applied at harvest on body weight was a combined (between and within) family selection approach. 81, 89, 96 and 114 families were produced for the Selection line in the F0, F1, F2 and F3 generations, respectively. In addition to the Selection line, 17 to 42 families were produced for the Control group in each generation. Results reported here are based on a data set consisting of 18,387 body and 1,730 carcass records, as well as full pedigree information collected over four generations. Variance and covariance components were estimated by restricted maximum likelihood fitting a multi-trait animal model. Experiments assessed performance of VIE tags in juvenile GFP of different size classes and individuals tagged with different numbers of tags showed that juvenile GFP at 2 g were of suitable size for VIE tags with no negative effects evident on growth or survival. Tag retention rates were above 97.8% and tag readability rates were 100% with a correct assignment rate of 95% through to mature animal size of up to 170 g. Across generations, estimates of heritability for body traits (body weight, body length, cephalothorax length, abdominal length, cephalothorax width and abdominal width) and carcass weight traits (abdominal weight, skeleton-off weight and telson-off weight) were moderate and ranged from 0.14 to 0.19 and 0.17 to 0.21, respectively. Body trait heritabilities estimated for females were significantly higher than for males whereas carcass weight trait heritabilities estimated for females and males were not significantly different (P > 0.05). Maternal and common environmental effects for body traits accounted for 4 to 5% of the total variance and were greater in females (7 to 10%) than in males (4 to 5%). Genetic correlations among body traits were generally high in both sexes. Genetic correlations between body and carcass weight traits were also high in the mixed sexes. Average selection response (% per generation) for body weight (transformed to square root) estimated as the difference between the Selection and the Control group was 7.4% calculated from least squares means (LSMs), 7.0% from estimated breeding values (EBVs) and 4.4% calculated from EBVs between two consecutive generations. Favourable correlated selection responses (estimated from LSMs) were detected for other body traits (12.1%, 14.5%, 10.4%, 15.5% and 13.3% for body length, cephalothorax length, abdominal length, cephalothorax width and abdominal width, respectively) over three selection generations. Data in the second selection generation showed positive correlated responses for carcass weight traits (8.8%, 8.6% and 8.8% for abdominal weight, skeleton-off weight and telson-off weight, respectively). Data in the third selection generation showed that heritability for body traits were moderate and ranged from 0.06 to 0.11 and 0.11 to 0.22 at weeks 10 and 18, respectively. Body trait heritabilities estimated at week 10 were not significantly lower than at week 18. Genetic correlations between body traits within age and genetic correlations for body traits between ages were generally high. Overall our results suggest that growth rate responds well to the application of family selection and carcass weight traits can also be improved in parallel, using this approach. Moreover, selection for high growth rate in GFP can be undertaken successfully before full market size has been reached. The outcome of this study was production of an improved culture strain of GFP for the Vietnamese culture industry that will be trialed in real farm production environments to confirm the genetic gains identified in the experimental stock improvement program.
Resumo:
Classifier selection is a problem encountered by multi-biometric systems that aim to improve performance through fusion of decisions. A particular decision fusion architecture that combines multiple instances (n classifiers) and multiple samples (m attempts at each classifier) has been proposed in previous work to achieve controlled trade-off between false alarms and false rejects. Although analysis on text-dependent speaker verification has demonstrated better performance for fusion of decisions with favourable dependence compared to statistically independent decisions, the performance is not always optimal. Given a pool of instances, best performance with this architecture is obtained for certain combination of instances. Heuristic rules and diversity measures have been commonly used for classifier selection but it is shown that optimal performance is achieved for the `best combination performance' rule. As the search complexity for this rule increases exponentially with the addition of classifiers, a measure - the sequential error ratio (SER) - is proposed in this work that is specifically adapted to the characteristics of sequential fusion architecture. The proposed measure can be used to select a classifier that is most likely to produce a correct decision at each stage. Error rates for fusion of text-dependent HMM based speaker models using SER are compared with other classifier selection methodologies. SER is shown to achieve near optimal performance for sequential fusion of multiple instances with or without the use of multiple samples. The methodology applies to multiple speech utterances for telephone or internet based access control and to other systems such as multiple finger print and multiple handwriting sample based identity verification systems.
Resumo:
Reliability of the performance of biometric identity verification systems remains a significant challenge. Individual biometric samples of the same person (identity class) are not identical at each presentation and performance degradation arises from intra-class variability and inter-class similarity. These limitations lead to false accepts and false rejects that are dependent. It is therefore difficult to reduce the rate of one type of error without increasing the other. The focus of this dissertation is to investigate a method based on classifier fusion techniques to better control the trade-off between the verification errors using text-dependent speaker verification as the test platform. A sequential classifier fusion architecture that integrates multi-instance and multisample fusion schemes is proposed. This fusion method enables a controlled trade-off between false alarms and false rejects. For statistically independent classifier decisions, analytical expressions for each type of verification error are derived using base classifier performances. As this assumption may not be always valid, these expressions are modified to incorporate the correlation between statistically dependent decisions from clients and impostors. The architecture is empirically evaluated by applying the proposed architecture for text dependent speaker verification using the Hidden Markov Model based digit dependent speaker models in each stage with multiple attempts for each digit utterance. The trade-off between the verification errors is controlled using the parameters, number of decision stages (instances) and the number of attempts at each decision stage (samples), fine-tuned on evaluation/tune set. The statistical validation of the derived expressions for error estimates is evaluated on test data. The performance of the sequential method is further demonstrated to depend on the order of the combination of digits (instances) and the nature of repetitive attempts (samples). The false rejection and false acceptance rates for proposed fusion are estimated using the base classifier performances, the variance in correlation between classifier decisions and the sequence of classifiers with favourable dependence selected using the 'Sequential Error Ratio' criteria. The error rates are better estimated by incorporating user-dependent (such as speaker-dependent thresholds and speaker-specific digit combinations) and class-dependent (such as clientimpostor dependent favourable combinations and class-error based threshold estimation) information. The proposed architecture is desirable in most of the speaker verification applications such as remote authentication, telephone and internet shopping applications. The tuning of parameters - the number of instances and samples - serve both the security and user convenience requirements of speaker-specific verification. The architecture investigated here is applicable to verification using other biometric modalities such as handwriting, fingerprints and key strokes.
Resumo:
This study reports the synthesis of extremely high aspect ratios (>3000) organic semiconductor nanowires of Ag–tetracyanoquinodimethane (AgTCNQ) on the surface of a flexible Ag fabric for the first time. These one-dimensional (1D) hybrid Ag/AgTCNQ nanostructures are attained by a facile, solution-based spontaneous reaction involving immersion of Ag fabrics in an acetonitrile solution of TCNQ. Further, it is discovered that these AgTCNQ nanowires show outstanding antibacterial performance against both Gram negative and Gram positive bacteria, which outperforms that of pristine Ag. The outcomes of this study also reflect upon a fundamentally important aspect that the antimicrobial performance of Ag-based nanomaterials may not necessarily be solely due to the amount of Ag+ ions leached from these nanomaterials, but that the nanomaterial itself may also play a direct role in the antimicrobial action. Notably, the applications of metal-organic semiconducting charge transfer complexes of metal-7,7,8,8-tetracyanoquinodimethane (TCNQ) have been predominantly restricted to electronic applications, except from our recent reports on their (photo)catalytic potential and the current case on antimicrobial prospects. This report on growth of these metal-TCNQ complexes on a fabric not only widens the window of these interesting materials for new biological applications, it also opens the possibilities for developing large-area flexible electronic devices by growing a range of metal-organic semiconducting materials directly on a fabric surface.
Resumo:
iTRAQ (isobaric tags for relative or absolute quantitation) is a mass spectrometry technology that allows quantitative comparison of protein abundance by measuring peak intensities of reporter ions released from iTRAQ-tagged peptides by fragmentation during MS/MS. However, current data analysis techniques for iTRAQ struggle to report reliable relative protein abundance estimates and suffer with problems of precision and accuracy. The precision of the data is affected by variance heterogeneity: low signal data have higher relative variability; however, low abundance peptides dominate data sets. Accuracy is compromised as ratios are compressed toward 1, leading to underestimation of the ratio. This study investigated both issues and proposed a methodology that combines the peptide measurements to give a robust protein estimate even when the data for the protein are sparse or at low intensity. Our data indicated that ratio compression arises from contamination during precursor ion selection, which occurs at a consistent proportion within an experiment and thus results in a linear relationship between expected and observed ratios. We proposed that a correction factor can be calculated from spiked proteins at known ratios. Then we demonstrated that variance heterogeneity is present in iTRAQ data sets irrespective of the analytical packages, LC-MS/MS instrumentation, and iTRAQ labeling kit (4-plex or 8-plex) used. We proposed using an additive-multiplicative error model for peak intensities in MS/MS quantitation and demonstrated that a variance-stabilizing normalization is able to address the error structure and stabilize the variance across the entire intensity range. The resulting uniform variance structure simplifies the downstream analysis. Heterogeneity of variance consistent with an additive-multiplicative model has been reported in other MS-based quantitation including fields outside of proteomics; consequently the variance-stabilizing normalization methodology has the potential to increase the capabilities of MS in quantitation across diverse areas of biology and chemistry.
Resumo:
Robust descriptor matching across varying lighting conditions is important for vision-based robotics. We present a novel strategy for quantifying the lighting variance of descriptors. The strategy works by utilising recovered low dimensional mappings from Isomap and our measure of the lighting variance of each of these mappings. The resultant metric allows different descriptors to be compared given a dataset and a set of keypoints. We demonstrate that the SIFT descriptor typically has lower lighting variance than other descriptors, although the result depends on semantic class and lighting conditions.
Resumo:
This paper proposes techniques to improve the performance of i-vector based speaker verification systems when only short utterances are available. Short-length utterance i-vectors vary with speaker, session variations, and the phonetic content of the utterance. Well established methods such as linear discriminant analysis (LDA), source-normalized LDA (SN-LDA) and within-class covariance normalisation (WCCN) exist for compensating the session variation but we have identified the variability introduced by phonetic content due to utterance variation as an additional source of degradation when short-duration utterances are used. To compensate for utterance variations in short i-vector speaker verification systems using cosine similarity scoring (CSS), we have introduced a short utterance variance normalization (SUVN) technique and a short utterance variance (SUV) modelling approach at the i-vector feature level. A combination of SUVN with LDA and SN-LDA is proposed to compensate the session and utterance variations and is shown to provide improvement in performance over the traditional approach of using LDA and/or SN-LDA followed by WCCN. An alternative approach is also introduced using probabilistic linear discriminant analysis (PLDA) approach to directly model the SUV. The combination of SUVN, LDA and SN-LDA followed by SUV PLDA modelling provides an improvement over the baseline PLDA approach. We also show that for this combination of techniques, the utterance variation information needs to be artificially added to full-length i-vectors for PLDA modelling.
Resumo:
Recently, mean-variance analysis has been proposed as a novel paradigm to model document ranking in Information Retrieval. The main merit of this approach is that it diversifies the ranking of retrieved documents. In its original formulation, the strategy considers both the mean of relevance estimates of retrieved documents and their variance. How- ever, when this strategy has been empirically instantiated, the concepts of mean and variance are discarded in favour of a point-wise estimation of relevance (to replace the mean) and of a parameter to be tuned or, alternatively, a quantity dependent upon the document length (to replace the variance). In this paper we revisit this ranking strategy by going back to its roots: mean and variance. For each retrieved document, we infer a relevance distribution from a series of point-wise relevance estimations provided by a number of different systems. This is used to compute the mean and the variance of document relevance estimates. On the TREC Clueweb collection, we show that this approach improves the retrieval performances. This development could lead to new strategies to address the fusion of relevance estimates provided by different systems.
Resumo:
Background: Physical activity after breast cancer diagnosis is associated with improved survival. This study examines levels of and changes in physical activity following breast cancer diagnosis, overall and by race. Methods: The Carolina Breast Cancer Study, Phase III, assessed pre- and post-diagnosis physical activity levels in a cohort of 1,735 women, aged 20-74, diagnosed with invasive breast cancer between 2008 and 2011 in 44 counties of North Carolina. Logistic regression and analysis of variance were used to examine whether demographic, behavioral and clinical characteristics were associated with activity levels. Results: Only 35% of breast cancer survivors met current physical activity guidelines post-diagnosis. A decrease in activity following diagnosis was reported by 59% of patients, with the average study participant reducing their activity by 230 minutes (95% CI: 190, 270). Following adjustment for potential confounders, when compared to white women, African-American women were less likely to meet national physical activity guidelines post-diagnosis (odds ratio: 1.38, 95% CI: 1.01, 1.88), reported less weekly post-diagnosis physical activity (182 vs. 215 minutes; p=0.13), and reported higher average reductions in pre- versus post-diagnosis weekly activity (262 vs. 230 minutes; p-value = 0.13). Conclusion: Despite compelling evidence demonstrating the benefits of physical activity post-breast cancer, it is clear that more work needs to be done to promote physical activity in breast cancer patients, especially among African-American women.
Resumo:
Integer ambiguity resolution is an indispensable procedure for all high precision GNSS applications. The correctness of the estimated integer ambiguities is the key to achieving highly reliable positioning, but the solution cannot be validated with classical hypothesis testing methods. The integer aperture estimation theory unifies all existing ambiguity validation tests and provides a new prospective to review existing methods, which enables us to have a better understanding on the ambiguity validation problem. This contribution analyses two simple but efficient ambiguity validation test methods, ratio test and difference test, from three aspects: acceptance region, probability basis and numerical results. The major contribution of this paper can be summarized as: (1) The ratio test acceptance region is overlap of ellipsoids while the difference test acceptance region is overlap of half-spaces. (2) The probability basis of these two popular tests is firstly analyzed. The difference test is an approximation to optimal integer aperture, while the ratio test follows an exponential relationship in probability. (3) The limitations of the two tests are firstly identified. The two tests may under-evaluate the failure risk if the model is not strong enough or the float ambiguities fall in particular region. (4) Extensive numerical results are used to compare the performance of these two tests. The simulation results show the ratio test outperforms the difference test in some models while difference test performs better in other models. Particularly in the medium baseline kinematic model, the difference tests outperforms the ratio test, the superiority is independent on frequency number, observation noise, satellite geometry, while it depends on success rate and failure rate tolerance. Smaller failure rate leads to larger performance discrepancy.
Resumo:
We examine some variations of standard probability designs that preferentially sample sites based on how easy they are to access. Preferential sampling designs deliver unbiased estimates of mean and sampling variance and will ease the burden of data collection but at what cost to our design efficiency? Preferential sampling has the potential to either increase or decrease sampling variance depending on the application. We carry out a simulation study to gauge what effect it will have when sampling Soil Organic Carbon (SOC) values in a large agricultural region in south-eastern Australia. Preferential sampling in this region can reduce the distance to travel by up to 16%. Our study is based on a dataset of predicted SOC values produced from a datamining exercise. We consider three designs and two ways to determine ease of access. The overall conclusion is that sampling performance deteriorates as the strength of preferential sampling increases, due to the fact the regions of high SOC are harder to access. So our designs are inadvertently targeting regions of low SOC value. The good news, however, is that Generalised Random Tessellation Stratification (GRTS) sampling designs are not as badly affected as others and GRTS remains an efficient design compared to competitors.
Resumo:
The formation of vertically aligned single-crystalline silicon nanostructures via "self-organized" maskless etching in Ar+ H 2 plasmas is studied. The shape and aspect ratio can be effectively controlled by the reactive plasma composition. In the optimum parameter space, single-crystalline pyramid-like nanostructures are produced; otherwise, nanocones and nanodots are formed. This generic nanostructure formation approach does not involve any external material deposition. It is based on a concurrent sputtering, etching, hydrogen termination, and atom/radical redeposition and can be applied to other nanomaterials.