Linear discriminate analysis

Purpose This study was undertaken to identify and compare the prognostic value of gene expression, chromosomal, and clinico-pathological data for the prediction of subsequent metastases in patients with primary uveal melanoma. Patients and Methods For comparison of different sets of predictor variables diagonal linear discriminant analysis was used. Chromosomal events were assessed by comparative genomic hybridization and gene expression profiling by microarray. Twenty-eight patients with a median follow-up of 68 months were analyzed, of whom 12 had developed subsequent metastases. Results Diagonal linear discriminant analysis with crossvalidation of gene expression data detected 42 genes as differentially expressed in metastasizing vs non-metastasizing uveal melanomas in all 28 cases. Comparing quantitative scores of discriminant analysis, grouping precision was significant better with gene expression profiling compared to comparative genomic hybridization (P ¼ 0.01) and to clinical data (P ¼ 0.001). Two published gene lists associated with monosomy 3 and metastatic tumor growth were used as classifier for discriminant analysis and yielded superior classification in patients with and without subsequent metastases than chromosomal or clinico-pathological data. Conclusion In our patient cohort gene expression profiling of primary uveal melanoma tissue was superior to clinical-pathological and chromosomal analysis to assess for the risk of subsequent metastases.

We make use of discrete wavelets to extract distinguishing features between normal and cancerous human breast tissue fluorescence spectra. These are then used in conjunction with discriminant analysis for the purpose of reliable tissue differentiation. The wavelet coefficients at different levels of decomposition, representing intensity variations at different scales, are selected as feature vectors wherein the multiresolution and localization properties of wavelets are optimally exploited for identifying features. This wavelet based approach, when combined with the sensitive polarized fluorescence data, yielded statistically reliable characterization of tissue types for diagnostic purpose. Analysis of a number of data sets belonging to both perpendicular and parallel polarized spectra have led to key distinctions between cancerous, benign and normal tissues. of computer aided diagnostics (CAD) . Amongst various optical methods for tissue diagnostics, fluorescence spectroscopy [2] is one of the preferred choices, due to its sensitivity to molecular environment and biochemical changes that take place during the progress of disease. A number of key fluorophores have been identified, which show marked differences in behavior, in cancerous and normal tissues. Significant progress in cancer diagnosis using tissue fluorescence has been achieved, since the early work of Alfano and coworkers, on the tumor detection in rat tissues . Many of these native fluorophores used for laserinduced fluorescence need to be excited through ultraviolet light; for example UV excited fluorescence probes NADH, tryptophan and tyrosine. In this work, we consider the flourescence due to fluorophores like flavin and porphyrin, which can be excited by visible light, thereby avoiding the potentially damaging effect of UV light. Key words -Cancer diagnostics, discrete wavelet transform, high-pass coefficients, linear discriminant analysis.

Diffuse subtypes of cutaneous lymphoid hyperplasia (CLH; = 18) and primary malignant follicular center cell lym-J1 homa of the skin (FCCL, n = 11) were diagnosed by COI1 -~cntional histology, immunophenotyping on paraffin sections, and gene rearran~ement a~a l ysis. We. then counted on scmithin, Azur A -sta1l1ed. sectlOns of res1l1-re-e~bedd.ed biopsy specimens the relauve numbers of apoptotlC bodies Jllong all lymphoid cells (apoptotic index [AI]). The diag-~ostic value of AI was compared to that of mitotic indices (MI) and percentages of various cell types in the cutaneous •IJiltrate. Features of cellular infiltrates distinguishing to two ~roups of lesions, in the order of decreasing significance, were percent large ly~pho. id cells, percent medium-sized lymphoid cells (both higher 111 FCCL); percent small lymphoid cells, percent epithelioid/giant ceJls, and percent histio-T he differential diagnosis of benign and malignantcutaneous lymphoid cell infiltrates is one of the most difficult tasks in dermatopathology . Immunohistochemistry has greatly helped to distinguish cutaneous lymphoid hyperplasia (CLH), also called pseudolymho ma or lymphadenosis benigna cutis, from malignant lympho-~a5 other than mycosis fungoides [2,3]; however, markers are not infallib le . Further, there ~re pit~alls in the application of molecular biologic methods, e.g., 111 the lI1terpretauon of Sou.thern blots . Thus, the e.valuatioll of doubtful cases wIth lymphOId cellllllilcrates in the skl11 must sull be based 01: a constell~tlon of features, including morphology, and not on a sll1gle wtenon .

Data projection is a commonly used technique applied to analyse high dimensional data. In the present work, we propose a new data projection method that uses genetic algorithms to find linear projections, providing meaningful representations of the original data. The proposed technique is compared with well known methods as Principal Components Analysis (PCA) and neural networks for non-linear discriminant analysis (NDA). A comparative study of these methods with several data sets is presented.

Background: Assessment of bladder outlet obstruction (BOO) is standard clinical practice in patients with lower urinary tract symptoms (LUTS). This is currently achieved through pressure-flow studies. Research indicates that progressive functional impairment of the bladder due to BOO is associated with haemodynamic changes. Near-infrared spectroscopy (NIRS) is an optical method of monitoring tissue oxygenation and haemodynamics via changes in concentration of the chromophores oxyhaemoglobin (O 2 Hb) and deoxyhaemoglobin (HHb). Objective: To report a noninvasive technique and mathematic method of analysis for assessment of BOO in male subjects using NIRS and to test the independent ability of NIRS data to distinguish between patients with and without obstruction using a classification and regression tree algorithm (CART). Design, setting, and participants: A prospective cohort study to evaluate subjects presenting for urodynamic assessment of LUTS using standard urodynamic studies with simultaneous transcutaneous NIRS monitoring. The NIRS data (magnitude and pattern of changes in O 2 Hb and HHb) were analysed, and a CART algorithm was constructed. Sixty-four males referred for evaluation of LUTS were studied at a tertiary care, university-based research and clinical facility. Measurements: Clinical symptoms were classified using the International Prostate Symptom Score (IPSS). Pressure-flow studies were done with simultaneous transcutaneous NIRS monitoring of the detrusor. Pressure-flow studies were classified according to the Abrams-Griffiths nomogram. NIRS data documented changes in the concentration of the chromophores O 2 Hb and HHb. Results and limitations: IPSS scores ranged from 12 to 34, with a mean of 19. The pressure-flow nomogram found 30 patients with BOO, 16 patients without BOO, and 18 patients with equivocal results. The CART found a misclassification error of 4% with 88% specificity and 94% precision. The NIRS instrument and algorithm were new; no asymptomatic subjects were studied. Conclusion: Using a CART algorithm, noninvasive NIRS data during voiding had independent discriminatory ability related to classification of BOO.

Mammalian antral oocytes with a Hoescht-positive DNA ring around the nucleolus (SN) are able to resume meiosis and to fully support the embryonic development, while oocytes with a non-surrounded nucleolus (NSN) cannot. Here, we applied FTIR microspectroscopy to characterize single SN and NSN mouse oocytes in order to try to elucidate some aspects of the mechanisms behind the different chromatin organization that impairs the full development of NSN oocyte-derived embryos. To this aim, oocytes were measured at three different stages of their maturation: just after isolation and classification as SN and NSN oocytes (time 0); after 10 h of in vitro maturation, i.e. at the completion of the metaphase I (time 1); and after 20 h of in vitro maturation, i.e. at the completion of the metaphase II (time 2). Significant spectral differences in the lipid (3050-2800 cm -1 ) and protein (1700-1600 cm -1 ) absorption regions were found between the two types of oocytes and among the different stages of maturation within the same oocyte type. Moreover, dramatic changes in nucleic acid content, concerning mainly the extent of transcription and polyadenylation, were detected in particular between 1000 and 800 cm -1 . The use of the multivariate principal component-linear discriminant analysis (PCA-LDA) enabled us to identify the maturation stage in which the separation between the two types of oocytes took place, finding as the most discriminating wavenumbers those associated to transcriptional activity and polyadenylation, in agreement with the visual analysis of the spectral data.

Gilles.Hosch@field.fao.org Abstract Traditional subsistence and artisanal fisheries provide sustenance, employment and income to vast numbers of Pacific islanders, and are of great importance to national economies and food security. The pervasive lack of subsistence fisheries data across the region and the worrying number of accounts of collapsing stocks and species extinctions are of growing concern to authorities involved in the management of coastal fishery resources. The author explores new grounds on how to overcome two fundamental hurdles for the successful management of subsistence fisheries. It is argued that under a scenario of increasing fishing pressures and changing societies, basic data to found management decisions upon, and awareness of the new generation towards the vulnerability of their resources are fundamental. The proposed way of achieving both is through the "Artisanal Fisheries Student Census". Secondary schools integrate an assignment into their sci...

In this report, we propose the combination of computational methods and in vivo primary screening in zebrafish larvae and confirmatory in mice models as a novel strategy to accelerate anti-inflammatory drug discovery. Initially, a database of 1213 organic chemicals with great structural variability e 587 of them anti-inflammatory agents plus 626 compounds with other clinical uses e was divided into training and test groups. Atom-based quadratic indices e a TOMOCOMD-CARDD molecular descriptors family e and linear discriminant analysis (LDA) were used to develop a total of 13 models to describe the anti-inflammatory activity. The best model (Eq. ( )) shows an accuracy of 87.70% in the training set, and values of Matthews correlation coefficient (C) of 0.75. The robustness of the models was demonstrated using an external test set as validation method, i.e., Eq. ( ) revealing classification of 88.44% (C ¼ 0.77) in this series. All models were employed to develop ensemble a QSAR classification system, in which the individual QSAR outputs are the inputs of the aforementioned fusion approach. The fusion model was used for the identification of novel anti-inflammatory compounds using virtual screening of 145 molecules available in our in-house library of indazole, indole, cinnoline and quinoxaline derivatives. Out of these, 34 chemicals were selected, synthesized and tested in a lipopolysaccharide (LPS)-induced leukocyte migration assay in zebrafish larvae. This activity was evaluated based on leukocyte migration to the injury zone of tail-transected larvae. Compounds 18 (3 mM), 24 (10 mM), 25 (10 mM), 6 (10 mM), 15 (30 mM), 11 (30 mM) and 12 (30 mM) gave the best results displaying relative leukocyte migration (RLM)

The influence of seven variables on the results of zeolite synthesis performed in presence of 1-methyl-1,4-tetramethylenbis(N-methylpiperidinium)-dihydroxide as structure directing agent (SDA) was examined. The syntheses were planned according to a factorial design strategy and run by using a multiautoclave device. The produced data-set (experimental factors values and qualitative XRD responses for 72 syntheses) was explored by stepwise linear discriminant analysis (LDA): SiO 2 /Al 2 O 3 molar ratio was identified as the most influent synthesis variable, discriminating between a region rich in amorphous materials and a region where the formation of MFI zeolite is favoured.

This study constitutes a compr ehensive signal analysis appr oach to the mor phological char acter ization of mammogr aphic mass shape. Thr ee distinct areas of shape mor phology wer e exploited for featur e extr action. Specifically, the r adial distance signal, the DFT spectr um envelope and the DWT decomposition with multiple wavelet function choices, were analyzed by seven cur ve featur e functions, as carr ier s of significant discriminating infor mation. Classification was conducted against the mor phological shape type identification, as well as the ver ified clinical diagnosis, using optimized featur e set selections and combinations by multivar iate statistical significance analysis. All available datasets and configur ations wer e applied to a wide r ange of linear and neur al classifiers, including linear discriminant analysis, leastsquar es minimum distance, Knear est neighbor , RBF and MLP neur al networks. Neur al classifier s outper for med linear equivalents in all cases, pr oducing an over all accur acy of 72.3% for mor phological shape type identification and 89.2% for clinical diagnosis identification.

This off-line study aims to assess the performance of five classifiers commonly used in the brain-computer interface (BCI) community, when applied to a gaze-independent P300-based BCI. In particular, we compared the results of four linear classifiers and one nonlinear: Fisher's linear discriminant analysis (LDA), stepwise linear discriminant analysis (SWLDA), Bayesian linear discriminant analysis (BLDA), linear support vector machine (LSVM) and Gaussian supported vector machine (GSVM). Moreover, different values for the decimation of the training dataset were tested. The results were evaluated both in terms of accuracy and written symbol rate with the data of 19 healthy subjects. No significant differences among the considered classifiers were found. The optimal decimation factor spanned a range from 3 to 24 (12 to 94 ms long bins). Nevertheless, performance on individually optimized classification parameters is not significantly different from a classification with general parameters (i.e. using an LDA classifier, about 48 ms long bins).

[1] apply the "kernel trick" to obtain a non-linear variant of Fisher's linear discriminant analysis method, demonstrating state-of-the-art performance on a range of benchmark datasets. We show that leave-one-out cross-validation of kernel Fisher discriminant classifiers can be implemented with a computational complexity of only O( 3 ) operations rather than the O( 4 ) of a naïve implementation, where is the number of training patterns. Leave-one-out cross-validation then becomes an attractive means of model selection in large-scale applications of kernel Fisher discriminant analysis, being significantly faster than conventional k-fold cross-validation procedures commonly used.

Implanted microelectrode arrays can directly pick up electrode signals from the primary motor cortex (M1) during movement, and brain-machine interfaces (BMIs) can decode these signals to predict the directions of contemporaneous movements. However, it is not well known how much each individual input is responsible for the overall performance of a BMI decoder. In this paper, we seek to quantify how much each channel contributes to an artificial neural network (ANN)-based decoder, by measuring how much the removal of each individual channel degrades the accuracy of the output. If information on movement direction was equally distributed among channels, then the removal of one would have a minimal effect on decoder accuracy. On the other hand, if that information was distributed sparsely, then the removal of specific information-rich channels would significantly lower decoder accuracy. We found that for most channels, their removal did not significantly affect decoder performance. However, for a subset of channels (16 out of 61), removing them significantly reduced the decoder accuracy. This suggests that information is not uniformly distributed among the recording channels. We propose examining these channels further to optimize BMIs more effectively, as well as understand how M1 functions at the neuronal level.

This work serves as an initial investigation into improvements to classification accuracy of an imagined movement-based Brain Computer Interface (BCI) by combining the feature spaces of two unique measurement modalities: functional near infrared spectroscopy (fNIRS) and electroencephalography (EEG). Our dual-modality system recorded concurrent and co-locational hemodynamic and electrical responses in the motor cortex during an imagined movement task, participated in by two subjects. Offline analysis and classification of fNIRS and EEG data was performed using leave-one-out cross-validation (LOOCV) and linear discriminant analysis (LDA). Classification of 2dimensional fNIRS and EEG feature spaces was performed separately and then their feature spaces were combined for further classification. Results of our investigation indicate that by combining feature spaces, modest gains in classification accuracy of an imagined movement-based BCI can be achieved by employing a supplemental measurement modality. It is felt that this technique may be particularly useful in the design of BCI devices for the augmentation of rehabilitation therapy.

QSAR (quantitative structureeactivity relationship) studies of tyrosinase inhibitors employing Dragon descriptors and linear discriminant analysis (LDA) are presented here. A data set of 653 compounds, 245 with tyrosinase inhibitory activity and 408 having other clinical uses were used. The active data set was processed by k-means cluster analysis in order to design training and prediction series. Seven LDA-based QSAR models were obtained. The discriminant functions applied showed a globally good classification of 99.79% for the best model Class ¼ À96:067 þ 1:988 Â 10 2 X0Av þ 91:907 BIC3 þ 6:853 CIC1 in the training set. External validation processes to assess the robustness and predictive power of the obtained model were carried out. This external prediction set had an accuracy of 99.44%. After that, the developed models were used in ligand-based virtual screening of tyrosinase inhibitors from the literature and never considered in either training or predicting series. In this case, all screened chemicals were correctly classified by the LDA-based QSAR models. As a final point, these fitted models were used in the screening of new bipiperidine series as new tyrosinase inhibitors. These methods are an adequate alternative to the process of selection/identification of new bioactive compounds. The biosilico assays and in vitro results of inhibitory activity on mushroom tyrosinase showed good correspondence. It is important to stand out that compound BP4 (IC 50 ¼ 1.72 mM) showed higher activity in the inhibition against the enzyme than reference compound kojic acid (IC 50 ¼ 16.67 mM) and L-mimosine (IC 50 ¼ 3.68 mM). These results support the role of biosilico algorithm for the identification of new tyrosinase inhibitor compounds.

In this report, we propose the combination of computational methods and in vivo primary screening in zebrafish larvae and confirmatory in mice models as a novel strategy to accelerate anti-inflammatory drug discovery. Initially, a database of 1213 organic chemicals with great structural variability e 587 of them anti-inflammatory agents plus 626 compounds with other clinical uses e was divided into training and test groups. Atom-based quadratic indices e a TOMOCOMD-CARDD molecular descriptors family e and linear discriminant analysis (LDA) were used to develop a total of 13 models to describe the anti-inflammatory activity. The best model (Eq. ( )) shows an accuracy of 87.70% in the training set, and values of Matthews correlation coefficient (C) of 0.75. The robustness of the models was demonstrated using an external test set as validation method, i.e., Eq. ( ) revealing classification of 88.44% (C ¼ 0.77) in this series. All models were employed to develop ensemble a QSAR classification system, in which the individual QSAR outputs are the inputs of the aforementioned fusion approach. The fusion model was used for the identification of novel anti-inflammatory compounds using virtual screening of 145 molecules available in our in-house library of indazole, indole, cinnoline and quinoxaline derivatives. Out of these, 34 chemicals were selected, synthesized and tested in a lipopolysaccharide (LPS)-induced leukocyte migration assay in zebrafish larvae. This activity was evaluated based on leukocyte migration to the injury zone of tail-transected larvae. Compounds 18 (3 mM), 24 (10 mM), 25 (10 mM), 6 (10 mM), 15 (30 mM), 11 (30 mM) and 12 (30 mM) gave the best results displaying relative leukocyte migration (RLM)

Existing Trichomonas vaginalis therapies are out of reach for most trichomoniasis people in developing countries and, where available, they are limited by their toxicity (mainly in pregnant women) and their cost. New antitrichomonal agents are needed to combat emerging metronidazole-resistant trichomoniasis and reduce the side effects associated with currently available drugs. Toward this end, atom-based bilinear indices, a new TOMOCOMD-CARDD molecular descriptor, and linear discriminant analysis (LDA) were used to discover novel, potent, and non-toxic lead trichomonacidal chemicals. Two discriminant functions were obtained with the use of non-stochastic and stochastic atom-type bilinear indices for heteroatoms and H-bonding of heteroatoms. These atomic-level molecular descriptors were calculated using a weighting scheme that includes four atomic labels, namely atomic masses, van der Waals volumes, atomic polarizabilities, and atomic electronegativities in Pauling scale. The obtained LDA-based QSAR models, using non-stochastic and stochastic indices, were able to classify correctly 94.51% (90.63%) and 93.41% (93.75%) of the chemicals in training (test) sets, respectively. They showed large Matthews' correlation coefficients (C); 0.89 (0.79) and 0.87 (0.85), for the training (test) sets, correspondingly. The result of predictions on the 15% full-out cross-validation test also evidenced the robustness and predictive power of the obtained models. In addition, canonical regression analyses corroborated the statistical quality of these models (R can of 0.749 and of 0.845, correspondingly); they were also used to compute biological activity canonical scores for each compound. On the other hand, a close inspection of the molecular descriptors included in both equations showed that several of these molecular fingerprints are strongly interrelated with each other. Therefore, these models were orthogonalized using the Randic ´orthogonalization procedure. These classification functions were then applied to find new lead antitrichomonal agents and six compounds were selected as possible active compounds by computational screening. The designed compounds were synthesized and tested for in vitro activity against T. vaginalis. Out of the six compounds that were designed, and synthesized, three molecules (chemicals VA5-5a, VA5-5c, and VA5-12b) showed high to moderate cytocidal activity at the

Subspace selection is a powerful tool in data mining. An important subspace method is the Fisher-Rao linear discriminant analysis (LDA), which has been successfully applied in many fields such as biometrics, bioinformatics, and multimedia retrieval. However, LDA has a critical drawback: the projection to a subspace tends to merge those classes that are close together in the original feature space. If the separated classes are sampled from Gaussian distributions, all with identical covariance matrices, then LDA maximizes the mean value of the Kullback-Leibler (KL) divergences between the different classes. We generalize this point of view to obtain a framework for choosing a subspace by 1) generalizing the KL divergence to the Bregman divergence and 2) generalizing the arithmetic mean to a general mean. The framework is named the general averaged divergence analysis (GADA). Under this GADA framework, a geometric mean divergence analysis (GMDA) method based on the geometric mean is studied. A large number of experiments based on synthetic data show that our method significantly outperforms LDA and several representative LDA extensions.

To precisely model the time dependency of features is one of the important issues for speech recognition. Segmental unit input HMM with a dimensionality reduction method has been widely used to address this issue. Linear discriminant analysis (LDA) and heteroscedastic extensions, e.g., heteroscedastic linear discriminant analysis (HLDA) or heteroscedastic discriminant analysis (HDA), are popular approaches to reduce dimensionality. However, it is difficult to find one particular criterion suitable for any kind of data set in carrying out dimensionality reduction while preserving discriminative information. In this paper, we propose a new framework which we call power linear discriminant analysis (PLDA). PLDA can be used to describe various criteria including LDA, HLDA, and HDA with one control parameter. In addition, we provide an efficient selection method using a control parameter without training HMMs nor testing recognition performance on a development data set. Experimental results show that the PLDA is more effective than conventional methods for various data sets.

Background: Brain and nervous system tumours are the most common solid cancers in children. Molecular characterisation of these tumours is important for providing novel biomarkers of disease and identifying molecular pathways which may provide putative targets for new therapies. 1H magic angle spinning NMR spectroscopy (1H HR-MAS) is a powerful tool for determining metabolite profiles from small pieces of intact tissue and could potentially provide important molecular information. Methods: Forty tissue samples from 29 children with glial and primitive neuro-ectodermal tumours were analysed using HR-MAS (600 MHz Varian gHX nanoprobe). Tumour spectra were fitted to a library of individual metabolite spectra to provide metabolite values. These values were then used in a two tailed t-test and multi-variate analysis employing a principal component analysis and a linear discriminant analysis. Classification accuracy was estimated using a leave-one-out analysis and B632+ bootstrapping. Results: Glial tumours had significantly (two tailed t-test p < 0.05) higher creatine and glutamine and lower taurine, phosphoethanolamine, phosphorylcholine and choline compared with primitive neuro-ectodermal tumours. Classification accuracy was 90%. Medulloblastomas (n = 9) had significantly (two tailed t-test p < 0.05) higher creatine, glutamine, phosphorylcholine, glycine and scyllo-inositol than neuroblastomas (n = 7), classification accuracy was 94%. Supratentorial primitive neuro-ectodermal tumours had metabolite profiles in keeping with other primitive neuroectodermal tumours whilst ependymomas (n = 2) had metabolite profiles intermediate between pilocytic astrocytomas (n = 10) and primitive neuro-ectodermal tumours. HR-MAS identified key differences in the metabolite profiles of childhood brain and nervous system improving the molecular characterisation of these tumours. Further investigation of the underlying molecular pathways is required to assess their potential as targets for new agents.

Mel-frequency cepstral coefficients (MFCC) have been dominantly used in speaker recognition as well as in speech recognition. However, based on theories in speech production, some speaker characteristics associated with the structure of the vocal tract, particularly the vocal tract length, are reflected more in the high frequency range of speech. This insight suggests that a linear scale in frequency may provide some advantages in speaker recognition over the mel scale. Based on two state-of-theart speaker recognition back-end systems (one Joint Factor Analysis system and one Probabilistic Linear Discriminant Analysis system), this study compares the performances between MFCC and LFCC (Linear frequency cepstral coefficients) in the NIST SRE (Speaker Recognition Evaluation) 2010 extended-core task. Our results in SRE10 show that, while they are complementary to each other, LFCC consistently outperforms MFCC, mainly due to its better performance in the female trials. This can be explained by the relatively shorter vocal tract in females and the resulting higher formant frequencies in speech. LFCC benefits more in female speech by better capturing the spectral characteristics in the high frequency region. In addition, our results show some advantage of LFCC over MFCC in reverberant speech. LFCC is as robust as MFCC in the babble noise, but not in the white noise. It is concluded that LFCC should be more widely used, at least for the female trials, by the mainstream of the speaker recognition community.

High incidence of gastric cancer-related death is mainly due to diagnosis at an advanced stage in addition to the lack of adequate neoadjuvant therapy. Hence, new tools aimed at early diagnosis would have a positive impact in the outcome of the disease. Using cDNA arrays having 376 genes either identified previously as altered in gastric tumors or known to be altered in human cancer, we determined expression signature of 99 tissue fragments representing normal gastric mucosa, gastritis, intestinal metaplasia, and adenocarcinomas. We first validated the array by identifying molecular markers that are associated with intestinal metaplasia, considered as a transition stage of gastric adenocarcinomas of the intestinal type as well as markers that are associated with diffuse type of gastric adenocarcinomas. Next, we applied Fisher’s linear discriminant analysis in an exhaustive search of trios of genes that could be used to build classifiers for class distinction. Many classifiers could ...

The microbial community compositions of surface and subsurface marine sediments and sediments lining burrows of marine polychaetes and hemichordates from the North Inlet estuary (near Georgetown, S.C.) were analyzed by comparing ester-linked phospholipid fatty acid (PLFA) profiles with a back-propagating neural network (NN). The NNs were trained to relate PLFA inputs to sediment type outputs (e.g., surface, subsurface, and burrow lining) and worm species (e.g., Notomastus lobatus , Balanoglossus aurantiacus , and Branchyoasychus americana ). Sensitivity analysis was used to determine which of the 60 PLFAs significantly contributed to training the NN. The NN architecture was optimized by changing the number of hidden neurons and calculating the cross-validation error between predicted and actual outputs of training and test data. The optimal NN architecture was found to be four hidden neurons with 60-input neurons representing the 60 PLFAs, and four output neurons coding for both sed...

The classification of business cycles is a hard and important problem. Government as well as business decisions rely on the assessment of the current business cycle. In this paper, we investigate how economists can be better supported by a combination of machine learning techniques. We have successfully applied Inductive Logic Programming (ILP). For establishing time and value intervals different discretization procedures are discussed. The rule sets learned from different experiments were analyzed with respect to correlations in order to find a concept drift or shift.

The human T-cell leukemia virus type I (HTLV-I) is the causative agent for adult T-cell leukemia (ATL) and HTLV-I-associated myelopathy/tropical spastic paraparesis (HAM/TSP). Approximately 5% of infected individuals will develop either disease and currently there are no diagnostic tools for early detection or accurate assessment of disease state. We have employed high-throughput expression profiling of serum proteins using mass spectrometry to identify protein expression patterns that can discern between disease states of HTLV-I-infected individuals. Our study group consisted of 42 ATL, 50 HAM/TSP, and 38 normal controls. Spectral peaks corresponding to peptide ions were generated from MS-TOF data. We applied Classification and Regression Tree analysis to build a decision algorithm, which achieved 77% correct classification rate across the three groups. A second cohort of 10 ATL, 10 HAM and 10 control samples was used to validate this result. Linear discriminate analysis was performed to verify and visualize class separation. Affinity and sizing chromatography coupled with tandem mass spectrometry was used to identify three peaks specifically overexpressed in ATL: an 11.7 kDa fragment of alpha trypsin inhibitor, and two contiguous fragments (19.9 and 11.9 kDa) of haproglobin-2. To the best of our knowledge, this is the first application of protein profiling to distinguish between two disease states resulting from a single infectious agent.

Vitis vinifera L. cv Graciano is often used as a blending partner of Tempranillo based wines because it is considered to contribute significantly to the quality. The aim of this study is to discriminate between Tempranillo and Graciano monovarietal wines, and those made by the incorporation of 20% of Graciano variety in two different stages (at the end of malolactic fermentation and mixing the two grape varieties in the pre-fermentative maceration stage) of the winemaking process of the Tempranillo variety. To achieve this, supervised and unsupervised pattern recognition tools were applied to the data obtained in the study of the detailed polyphenolic composition, colour and other oenological parameters (143 variables). Patterns related to stages in the winemaking and ageing process, different wines and vintages can be observed using principal component analyses. Furthermore, linear discriminant analysis has been applied in order to characterise the wine samples. From the 143 variables, flavan-3-ols have exerted a profound influence on wine differentiation.

A sample set of 42 bronze Punic coins dating from the third century BC and excavated in the Phoenician, Punic and then Roman town of Tharros (western Sardinia) has been used to test the performance of the literature lead isotope database and to establish a well de®ned provenancing procedure. A selection from the available database, including Cypriot, Sardinian and Spanish ore deposits, has been used. The complete assignment has been carried out twice, using both statistical methods (linear discriminant analysis) and the conventional raw data plots. Three major and one minor disagreements have been found, suggesting that the two procedures, when properly employed, may be acceptably compatible. Almost one third of the coins exhibit a characteristic Cypriot signature, whereas only very few samples are compatible with Sardinian ores. However, the majority of the coins (c. 60%) cannot be associated with any known source, which suggests the possibility of extensive metal recycling.

Indication of cardiovascular risk is based on laboratory analysis of cardiovascular markers in blood. Assessment of diagnostic performance of traditionally used as well as new cardiovascular markers is the main goal of this paper. Performance of studied laboratory tests was evaluated by using the area below the corresponding receiver operating characteristic (ROC) curve. Application of an objective procedure for the determination of critical values of laboratory tests was based on maximum of efficiency. It was demonstrated that the patient's diagnosis can be verified or predicted not only by using the results of a chosen individual laboratory test but also by a multivariate statistical approach, which jointly utilizes all performed tests in the form of their optimal linear combination. An effective combination of original variables (laboratory tests) is provided either by principal component or discriminant function or logit-in case of logistic regression.

In this paper, we consider the problem of building models that have high subjectivity classification accuracy across domains. For that purpose, we present and evaluate new methods based on multi-view learning using both high-level (i.e. linguistic features for subjectivity detection) and lowlevel features (i.e. unigrams and bigrams). In particular, we show that multi-view learning, combining high-level and low-level features with adapted classifiers, can lead to improved results compared to one of the state-of-the-art algorithms called Stochastic Agreement Regularization. In particular, the experiments show that dividing the set of characteristics into three views returns the best results overall with accuracy across domains of 91.3% for the Class-Guided Multi-View Learning Algorithm, which combines both Linear Discriminant Analysis and Support Vector Machines.

State of vigilance is determined by behavioral observations and electrophysiological activity. Here, we improve automatic state of vigilance discrimination by combining head acceleration with EEG measures. We incorporated biaxial DC-sensitive microelectromechanical system (MEMS) accelerometers into head-mounted preamplifiers in rodents. Epochs (15 s) of behavioral video and EEG data formed training sets for the following states: Slow Wave Sleep, Rapid Eye Movement Sleep, Quiet Wakefulness, Feeding or Grooming, and Exploration. Multivariate linear discriminant analysis of EEG features with and without accelerometer features was used to classify behavioral state. A broad selection of EEG feature sets based on recent literature on state discrimination in rodents was tested. In all cases, inclusion of head acceleration significantly improved the discriminative capability. Our approach offers a novel methodology for determining the behavioral context of EEG in real time, and has potential application in automatic sleep-wake staging and in neural prosthetic applications for movement disorders and epileptic seizures.

The Indonesian SAR robot contest (KRSRI) is a development of the fire extinguisher robot contest (KRPAI); initially, the robot at KRPAI only put out fires. Still, at KRSRI, the robot was asked to prioritize the SAR function. The robot had to overcome obstacles in this contest to complete it. Based on this, an obstacle detection system for the robot was designed using machine learning with the K-nearest neighbor algorithm and gray level co-occurrence matrix feature extraction. Later, the robot is expected to be able to carry out accurate obstacle detection to prioritize efficiency so that no more time is consumed due to the robot incorrectly detecting an obstacle. The results of the tests that have been carried out show that the detection accuracy based on the test dataset is 80% for rising barriers, 100% for debris obstacles, and 90% for step obstacles, and an error value of 20% for increasing obstacles is obtained, 0% for debris obstacles, and 10% for stair obstacles.

Currently, genome-wide association studies (GWAS) are conducted by collecting a massive number of SNPs (i.e., large p) for a relatively small number of individuals (i.e., small n) and associations are made between clinical phenotypes and genetic variation one single-nucleotide polymorphism (SNP) at a time. Univariate association approaches like this ignore the linkage disequilibrium between SNPs in regions of low recombination. This results in a low reliability of candidate gene identification. Here we propose to improve the case-control GWAS approach by implementing linear discriminant analysis (LDA) through a penalized orthogonal-components regression (POCRE), a newly developed variable selection method for large p small n data. The proposed POCRE-LDA method was applied to the Genetic Analysis Workshop 16 case-control data for rheumatoid arthritis (RA). In addition to the two regions on chromosomes 6 and 9 previously associated with RA by GWAS, we identified SNPs on chromosomes 10 and 18 as potential candidates for further investigation.

Production of Sherry wine vinegar encompasses a variable period of aging in wood. Changes in phenolic compounds during Sherry vinegar aging were studied. Samples were divided according to their aging period in wood (less than 2 years, more than 2 years) and phenolic compounds determined in a total of 38 samples by means of HPLC and Photodiode Array Detection. Results of Multiple Analysis of Variance (MANOVA) con®rms that there were signi®cant dierences among both groups. Good recalling rates were achieved in both Linear Discriminant Analysis LDA (mean=92.9%) and Arti®cial Neural Networks trained by Back Propagation (BPANN) (mean=90.4%) whilst prediction abilities are more discrete. Applying the forward selection of variables, six variables were selected according to the Wilks' lambda criterion: gallic acid, 5-hydroxymethylfuraldehyde, coumaroyltartaric acid glycoside, p-hydroxybenzoic acid, caeic acid and isoquercetrin. Three of these compounds (gallic acid, hydroxymethylfuraldehyde and coumaroyltartaric acid glycoside) were shown to have an increasing trend through most aging systems.

Dynamic PET (dPET) studies with 18 F-FDG were performed in patients with soft-tissue sarcomas who received neoadjuvant chemotherapy early in the course of therapy. The goal of the study was to evaluate the impact of early dPET studies and assess their value with regard to the therapy outcome using histopathologic data. Methods: The evaluation included 31 patients with nonmetastatic soft-tissue sarcomas, who were treated with neoadjuvant chemotherapy consisting of etoposide, ifosfamide, and doxorubicin. Patients were examined before the onset of therapy and after the completion of the second cycle. Histopathologic response served for reference and was available for 25 of 31 patients. Response was defined as less than 10% viable tumor tissue in the resected tumor tissue. The following parameters were retrieved from dPET studies: standardized uptake value (SUV); fractal dimension; 2-compartment model with computation of K 1 , k 2 , k 3 , and k 4 (unit, 1/min); fractional blood volume; and influx according to Patlak. Results: The mean SUV was 4.6 before therapy and 2.8 after 2 cycles. The mean influx was 0.059 before therapy and 0.043 after 2 cycles. The mean SUV was 3.9 in the responders and 5.5 in the nonresponders before therapy. After therapy, responders revealed a mean SUV of 2.5, whereas nonresponders had a mean SUV of 3.5. We used linear discriminant analysis to categorize the patients into 2 groups: response (n 5 12) and nonresponse (n 5 13). The correct classification rate of the responders (positive predictive value) was generally higher (.67%) than that for the nonresponders. Finally, the combined use of the 2 predictor variables, namely SUV and influx, of each study led to the highest accuracy of 83%. This combination was particularly useful for the prediction of responders (positive predictive value, 92%). The use of the percentage change in maximum SUV led to an accuracy of 58%. Conclusion: On the basis of these results, only a multiparameter analysis based on kinetic 18 F-FDG data of a baseline study and after 2 cycles is helpful for the early prediction of chemosensitivity in patients with soft-tissue sarcomas receiving neoadjuvant chemotherapy.

Independent component analysis (ICA) Principal component analysis (PCA) Linear discriminant analysis (LDA) Support vector machines (SVM) a b s t r a c t In this work, we proposed a versatile signal processing and analysis framework for Electroencephalogram (EEG). Within this framework the signals were decomposed into the frequency sub-bands using DWT and a set of statistical features was extracted from the sub-bands to represent the distribution of wavelet coefficients. Principal components analysis (PCA), independent components analysis (ICA) and linear discriminant analysis (LDA) is used to reduce the dimension of data. Then these features were used as an input to a support vector machine (SVM) with two discrete outputs: epileptic seizure or not. The performance of classification process due to different methods is presented and compared to show the excellent of classification process. These findings are presented as an example of a method for training, and testing a seizure prediction method on data from individual petit mal epileptic patients. Given the heterogeneity of epilepsy, it is likely that methods of this type will be required to configure intelligent devices for treating epilepsy to each individual's neurophysiology prior to clinical operation.

Symbiosis with mycorrhizal fungi substantially impacts secondary metabolism and defensive traits of colonised plants. In the present study, we investigated the influence of mycorrhization (Glomus intraradices) on inducible indirect defences against herbivores using the model legume Medicago truncatula. Volatile emission by mycorrhizal and non-mycorrhizal plants was measured in reaction to damage inflicted by Spodoptera spp. and compared to the basal levels of volatile emission by plants of two different cultivars. Emitted volatiles were recorded using closed-loop stripping and gas chromatography/mass spectrometry. The documented volatile patterns were evaluated using multidimensional scaling to visualise patterns and stepwise linear discriminant analysis to distinguish volatile blends of plants with distinct physiological status and genetic background. Volatile blends emitted by different cultivars of M. truncatula prove to be clearly distinct, whereas mycorrhization only slightly influenced herbivore-induced volatile emissions. Still, the observed differences were sufficient to create classification rules to distinguish mycorrhizal and non-mycorrhizal plants by the volatiles emitted. Moreover, the effect of mycorrhization turned out to be opposed in the two cultivars examined. Root symbionts thus seem to alter indirect inducible defences of M. truncatula against insect herbivores. The impact of this effect strongly depends on the genetic background of the plant and, hence, in part explains the highly contradictory results on tripartite interactions gathered to date.

The eigenface method that uses principal component analysis (PCA) has been the standard and popular method used in face recognition. This paper presents a PCA -memetic algorithm (PCA-MA) approach for feature selection. PCA has been extended by MAs where the former was used for feature extraction/dimensionality reduction and the latter exploited for feature selection. Simulations were performed over ORL and YaleB face databases using Euclidean norm as the classifier. It was found that as far as the recognition rate is concerned, PCA-MA completely outperforms the eigenface method. We compared the performance of PCA extended with genetic algorithm (PCA-GA) with our proposed PCA-MA method. The results also clearly established the supremacy of the PCA-MA method over the PCA-GA method. We further extended linear discriminant analysis (LDA) and kernel principal component analysis (KPCA) approaches with the MA and observed significant improvement in recognition rate with fewer features. This paper also compares the performance of PCA-MA, LDA-MA and KPCA-MA approaches.

A very well-known traditional approach in discriminant analysis is to use some linear (or nonlinear) combination of measurement variables which can enhance class separability. For instance, a linear (or a quadratic) classifier finds the linear projection (or the quadratic function) of the measurement variables that will maximize the separation between the classes. These techniques are very useful in obtaining good lower dimensional view of class separability. Fisher's discriminant analysis, which is primarily motivated by the multivariate normal distribution, uses the first-and second-order moments of the training sample to build such classifiers. These estimates, however, are highly sensitive to outliers, and they are not reliable for heavy-tailed distributions. This paper investigates two distribution-free methods for linear classification, which are based on the notions of statistical depth functions. One of these classifiers is closely related to Tukey's half-space depth, while the other is based on the concept of regression depth. Both these methods can be generalized for constructing nonlinear surfaces to discriminate among competing classes. These depth-based methods assume some finite-dimensional parametric form of the discriminating surface and use the distributional geometry of the data cloud to build the classifier. We use a few simulated and real data sets to examine the performance of these discriminant analysis tools and study their asymptotic properties under appropriate regularity conditions.

A very well-known traditional approach in discriminant analysis is to use some linear (or nonlinear) combination of measurement variables which can enhance class separability. For instance, a linear (or a quadratic) classifier finds the linear projection (or the quadratic function) of the measurement variables that will maximize the separation between the classes. These techniques are very useful in obtaining good lower dimensional view of class separability. Fisher's discriminant analysis, which is primarily motivated by the multivariate normal distribution, uses the first-and second-order moments of the training sample to build such classifiers. These estimates, however, are highly sensitive to outliers, and they are not reliable for heavy-tailed distributions. This paper investigates two distribution-free methods for linear classification, which are based on the notions of statistical depth functions. One of these classifiers is closely related to Tukey's half-space depth, while the other is based on the concept of regression depth. Both these methods can be generalized for constructing nonlinear surfaces to discriminate among competing classes. These depth-based methods assume some finite-dimensional parametric form of the discriminating surface and use the distributional geometry of the data cloud to build the classifier. We use a few simulated and real data sets to examine the performance of these discriminant analysis tools and study their asymptotic properties under appropriate regularity conditions.

Beside some quality parameters, forty four elements were determined by ICP-OES in 19 honey samples from different places of Hungary, in this study. These samples have different botanical origin. Acacia (Robinia pseudo-acacia), linden (Tilia sp.), rape (Brassica sp.), coriander (Coriandrum sativum), cherry (Prunus cerasus), lavender (Lavendula angustifoia) and multifloral honey. Ca, K, S, P, Mg and Na were the most abundant elements (896; 556; 60,5, 48,2; 31,5; 31,1 mg/kg, respectively). The B, Fe and Se content generally ranged from 2 to 8 mg/kg. In addition, we determined the electrical conductivity, -which is closely connected with element-, sugar-and moisture content.

Face recognition has become an attractive field in computer based application development in the last few decades. That is because of the wide range of areas they used in. And because of the wide variations of faces, face recognition from the database images, real data, capture images and sensor images is challenging problem and limitation. Image processing, pattern recognition and computer vision are relevant subjects to face recognition field. The innovation of new approaches of face authentication technologies is continuous subject to build much strong face recognition algorithms. In this work, to identify a face, there are three major strategies for feature extractions are discussed. Appearance-based and Modelbased methods and hybrid techniques as feature extractions are discussed. Also, review of major person recognition research the characteristics of good face authentication applications, Classification, Distance measurements and face databases are discussed while the final suggested methods are presented. This research has six sections organized as follow: Section one is the introduction. Section two is dedicated to applications related to face recognition. In Section three, face recognition techniques are presented by details. Then, classification types are illustrated in Section four. In section five, standard face databases are presented. Finally, in Section six, the conclusion is presented followed by the list of references.

The characterization of beer samples has a lot of interest because their composition can affect the taste and stability of beer and consumer health. Flame atomic absorption spectrometry was used to determine Fe, Mn, Zn, Cu, Mg, Ca and Al. Sodium and K were determined by flame atomic emission spectrometry. A sample preparation method was developed, based on treatment with HNO and H O in a microwave oven. This has many advantages 3 2 2 over the methods found in the literature. The combination of the results of atomic spectrometry and the spectrum obtained by molecular absorption spectrometry provides information on the inorganic and organic components of the samples. The application of chemometric techniques to chemical composition data could be extremely useful for food quality control. The metal concentrations, the molecular absorption spectrum, the pH and conductivity of each sample were subject to analysis of variance and linear discriminant analysis. Twenty-five different beer samples were used to differentiate and classify different types of beers.

In this study, an intelligent diagnosis system for diabetes on Linear Discriminant Analysis (LDA) and Adaptive Network Based Fuzzy Inference System (ANFIS): LDA-ANFIS is presented. The structure of this LDA-ANFIS intelligent system for diagnosis of diabetes is composed by two phases: The Linear Discriminant Analysis (LDA) phase and classificiation by using ANFIS classifier phase. In first phase, Linear Discriminant Analysis (LDA) is used to separate features variables between healthy and patient (diabetes) data. In second phase, the healthy and patient (diabetes) features obtained in first phase are given to inputs of ANFIS classifier. The correct diagnosis performance of the LDA-ANFIS intelligent system is calculated by using sensitivity and specificity analysis, classification accuracy and confusion matrix respectively. The classification accuracy of this LDA-ANFIS intelligent system was obtained about 84.61%.

Due to widespread use of the Internet, efficient management of multimedia databases has attracted many researchers. Variety of techniques including database indexing, classification and feature extraction are developed. In this study, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) methods that have been widely used for face recognition problem in literature, are integrated to MUVIS, a multimedia indexing and classification system with an effective indexing structure. PCA and LDA are integrated into MUVIS and their classification performance on a human passport photo database is compared with that of Gabor Filter that already exists in MUVIS.

Supervised classification or pattern recognition is a method to solve decision problems in Social Sciences. It is organized on the basis of specific sets of predictor variables and the existence of classes known a priori. Based on a training sample, its main objective is to construct a classification rule in order to predict the class to which a new object