Semisupervised Dimensionality Reduction and Classification Through Virtual Label Regression

Semisupervised dimensionality reduction has been attracting much attention as it not only utilizes both labeled and unlabeled data simultaneously, but also works well in the situation of out-of-sample. This paper proposes an effective approach of semisupervised dimensionality reduction through label propagation and label regression. Different from previous efforts, the new approach propagates the label information from labeled to unlabeled data with a well-designed mechanism of random walks, in which outliers are effectively detected and the obtained virtual labels of unlabeled data can be well encoded in a weighted regression model. These virtual labels are thereafter regressed with a linear model to calculate the projection matrix for dimensionality reduction. By this means, when the manifold or the clustering assumption of data is satisfied, the labels of labeled data can be correctly propagated to the unlabeled data; and thus, the proposed approach utilizes the labeled and the unlabeled data more effectively than previous work. Experimental results are carried out upon several databases, and the advantage of the new approach is well demonstrated.

Exploiting side information in locality preserving projection

Even if the class label information is unknown, side information represents some equivalence constraints between pairs of patterns, indicating whether pairs originate from the same class. Exploiting side information, we develop algorithms to preserve both the intra-class and inter-class local structures. This new type of locality preserving projection (LPP), called LPP with side information (LPPSI), preserves the data's local structure in the sense that the close, similar training patterns will be kept close, whilst the close but dissimilar ones are separated. Our algorithms balance these conflicting requirements, and we further improve this technique using kernel methods. Experiments conducted on popular face databases demonstrate that the proposed algorithm significantly outperforms LPP. Further, we show that the performance of our algorithm with partial side information (that is, using only small amount of pair-wise similarity/dissimilarity information during training) is comparable with that when using full side information. We conclude that exploiting side information by preserving both similar and dissimilar local structures of the data significantly improves performance.

The importance of use and end-of-life phases to the life cycle greenhouse gas (GHG) emissions of concrete : a review

Global climate change is one of the most significant environmental impacts at the moment. One central issue for the building and construction industry to address global climate change is the development of credible carbon labelling schemes for building materials. Various carbon labelling schemes have been developed for concrete due to its high contribution to global greenhouse gas (GHG) emissions. However, as most carbon labelling schemes adopt cradle-to-gate as system boundary, the credibility of the eco-label information may not be satisfactory because recent studies show that the use and end-of-life phases can have a significant impact on the life cycle GHG emissions of concrete in terms of carbonation, maintenance and rehabilitation, other indirect emissions, and recycling activities. A comprehensive review on the life cycle assessment of concrete is presented to holistically examine the importance of use and end-of-life phases to the life cycle GHG quantification of concrete. The recent published ISO 14067: Carbon footprint of products – requirements and guidelines for quantification and communication also mandates the use of cradle-to-grave to provide publicly available eco-label information when the use and end-of-life phases of concrete can be appropriately simulated. With the support of Building Information Modelling (BIM) and other simulation technologies, the contribution of use and end-of-life phases to the life cycle GHG emissions of concrete should not be overlooked in future studies.

Partially Supervised Neighbor Embedding for Example-Based Image Super-Resolution

Neighbor embedding algorithm has been widely used in example-based super-resolution reconstruction from a single frame, which makes the assumption that neighbor patches embedded are contained in a single manifold. However, it is not always true for complicated texture structure. In this paper, we believe that textures may be contained in multiple manifolds, corresponding to classes. Under this assumption, we present a novel example-based image super-resolution reconstruction algorithm with clustering and supervised neighbor embedding (CSNE). First, a class predictor for low-resolution (LR) patches is learnt by an unsupervised Gaussian mixture model. Then by utilizing class label information of each patch, a supervised neighbor embedding is used to estimate high-resolution (HR) patches corresponding to LR patches. The experimental results show that the proposed method can achieve a better recovery of LR comparing with other simple schemes using neighbor embedding.

Supervised Gaussian Process Latent Variable Model for Dimensionality Reduction

The Gaussian process latent variable model (GP-LVM) has been identified to be an effective probabilistic approach for dimensionality reduction because it can obtain a low-dimensional manifold of a data set in an unsupervised fashion. Consequently, the GP-LVM is insufficient for supervised learning tasks (e. g., classification and regression) because it ignores the class label information for dimensionality reduction. In this paper, a supervised GP-LVM is developed for supervised learning tasks, and the maximum a posteriori algorithm is introduced to estimate positions of all samples in the latent variable space. We present experimental evidences suggesting that the supervised GP-LVM is able to use the class label information effectively, and thus, it outperforms the GP-LVM and the discriminative extension of the GP-LVM consistently. The comparison with some supervised classification methods, such as Gaussian process classification and support vector machines, is also given to illustrate the advantage of the proposed method.

An empirical study on the influence of environmental labels on consumers

Purpose - Despite research studies indicating strong support for labelling information, uncertainty remains with respect to how labels influence consumers. This paper attempts to empirically investigate how consumers who differ in terms of environmentalism respond to labels.

Design/methodology/approach - The data were collected on an Australian sample using a structured questionnaire administered on the phone using quantum research (CATI) data collection services. A total of 155 questionnaires were completed and used for data analysis. The data were analysed using both descriptive measures and correlations between variables.

Findings - There appears to be a proportion of consumers that find product labels hard to understand. The research found that there are consumers who will buy green products even if they are lower in quality in comparison to alternative products, but would look for environmental information on labels. With respect to price sensitive green consumers, there appears to be a relationship between price sensitivity and 'always' reading labels as well as indicating that there is 'sufficient' information on product labels to make informed purchase decisions.

Research limitations/implications - It was beyond the scope of the research to account for some of the utilitarian approaches to interpretation or in-depth comprehension of label information. The sample size of 155, although selected using a probability method, may, to some extent, limit the overall accuracy of the results.

Practical implications - Offers some important information on different green consumer segments that would alert managers on how best to position environmental labels. Findings such as 'Satisfied with labels' correlates with 'Labels are accurate', which suggests that businesses need to provide a clear, accurate and easily legible label design to encourage satisfaction with the accuracy of content and the communication aspects of a label.

Originality/value - Contributes to better understanding of green customers purchase intentions and the usefulness of ecological product labels. It offers some insights and assistance to businesses in planning their green product/labelling strategies.

Legislação de rotulagem nutricional: instrumento de informação na promoção de escolhas alimentares - Natal-RN

Resolutions of the Board of numbers 359 and 360, of December 23, 2003, relating to Nutrition Labelling for Packaged Foods, establish quality standards and provide the education activities for health consumption, since one of the factors that enable the selection of healthier foods are the food labels as an important part in nutrition education. This is a descriptive study with a cross-sectional design. To calculate sample size, it was considered a margin of error of 20%, confidence level of 95% and prevalence of 52.5% for verifying nutritional information in a pilot study conducted in 2007. A total of 145 subjects were interviewed, resulting in 371 consumers in Natal, Rio Grande do Norte, Brazil, in order to determine prevalence of consulting nutritional information contained on food labels as a nutritional guideline for consumers and its association with sociodemographic variables as well as identify the intervention measures suggested by intervieweds so that this information can be better used to select healthy foods. Twenty-five of the 69 supermarkets belonging to the Supermarket Association of Rio Grande do Norte (ASSURN) were randomly selected. Data collection relied on interviews and extensive direct observation, using a semi-structured form composed of eight closed questions, some of which were multiple choice, and ten open questions. The chi-square test was performed for statistical analysis, using Statistical Package for Social Sciences (SPSS) 15.0 software. Label information most consulted was: expiry date (91.6%); product brand (49.4%); nutritional information (47.0%); zero trans fat (32.9%); zero sugar content (12.8%); zero fat content (3.0%); rich in fibers (2.7%); whether light or diet (30.4%); list of ingredients (16.8%); whether the product contained gluten or not (4.1%). When asked about the importance of nutritional information, 96.8% of the subjects responded important or very important ; of these 46.6% and 3.8% reported partially or totally understanding the information presented. It was found that 41.6% of the consumers consulted nutritional information for dietary reasons related to nontransmissible chronic diseases and 35.7% to be able to choose healthier foods. The data show a significant association between motivation to choose healthier foods and higher family income and schooling (p<0.0001). The intervention measures mentioned to make nutritional information better understood and used were: information and orientation about nutritional information, provided by qualified professionals in the supermarkets, the commercial establishment or the product manufacturer (73.9%) and media disclosure about the nature, importance and purpose of nutritional information (42.9%). In despite of communication noises the consumers use the nutrition claim for the nutritional guidance, showing association with some demographic variables. However, they desire the implementation of intervention measures that can be contextualized in the political construction of nutrition education to promote healthy food choices

Consumer interpretation of ready to drink orange juice and nectar labelling

The aim of this study was to evaluate consumers' use and interpretation of ready to drink orange juice and nectar label information and its influence on the purchase decision. One hundred and sixty-seven consumers of ready to drink orange juice and nectar were interviewed. The labels were analysed to evaluate their conformance to Brazilian legislation. The manufacturing and shelf life date were the information most often checked, followed by health related issues. Brand, price and flavour were the most important factors for purchase decision. Brand and flavour showed significant association with consumer age. For most interviewed, 'nectar', 'whole' and 'natural' or '100% natural' were not well understood; they were not in accordance with the Brazilian legislation. 'Nectar', 'whole' and 'natural' or '100% natural' received a positive interpretation, whereas 'reconstituted juice' was considered a negative expression. Nevertheless, none of the labels completely conformed to the specific nutritional labelling legislation. © 2013 Institute of Food Science and Technology.

Consumer attitude towards purchasing intent for ready to drink orange juice and nectar

Purpose: The purpose of this paper was to evaluate the impact of some labeling aspects on the consumer intent to purchase ready to drink orange juice and nectar. Design/methodology/approach: The influence of label information on the consumer intent to purchase was evaluated by conjoint analysis using a convenience sample (n=149). A factorial design with four characteristics, price, brand, information about the product and kind of beverage, was used. Three levels were established for brand and product information, and two for price and kind of beverage. Findings: Low price, product information and market leading brand had positive impact. No preservatives/natural was the information that most influenced consumer's purchase intent. The ideal label showed the leading brand, low price and information no preservatives/natural. These results could be useful for strategic planning of consumer instruction and have important implications for Brazilian orange juice manufactures. Originality/value: Although the most widely consumed beverages in Brazil are ready to drink orange juice and nectar, it was unexpected that consumers did not know the differences between them and that kind of beverage was not an important factor for the purchase decision. © Emerald Group Publishing Limited.

Composição nutricional e perfil de ácidos graxos de batatas chips e sanacks extrusados

Pós-graduação em Engenharia e Ciência de Alimentos - IBILCE

Análise isotópica, físico-química, centesimal e energética de água de coco

Pós-graduação em Agronomia (Energia na Agricultura) - FCA

Algoritmos de aprendizado semi-supervisionado baseados em grafos aplicados na bioinformática

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Fluid intake patterns: an epidemiological study among children and adolescents in Brazil

Background: Energy from liquids is one of the most important factors that could impact on the high prevalence of children and adolescents obesity around the world. There are few data on the liquid consumption in Brazil. The aim of this study is to evaluate the volume and quality of liquids consumed by Brazilian children and adolescents and to determine the proportion of their daily energy intake composed of liquids. Methods: A multicenter study was conducted in five Brazilian cities; the study included 831 participants between 3 and 17 years of age. A four-day dietary record specific to fluids was completed for each individual, and the volume of and Kcal from liquid intake were evaluated. The average number of Kcal in each beverage was determined based on label information, and the daily energy intake data from liquids were compared with the recommendations of the National Health Surveillance Agency (Agencia Nacional de Vigilancia Sanitaria-ANVISA), the Brazilian food regulation authority, according to each subject's age. Results: As the children aged, the volume of carbonated beverages that they consumed increased significantly, and their milk intake decreased significantly. For children between the ages of 3 and 10, milk and dairy products contributed the greatest daily number of Kcal from liquids. Sugar sweetened beverages which included carbonated beverages, nectars and artificial beverages, accounted for 37% and 45% of the total Kcal from liquid intake in the 3-to 6-year-old and 7- to 10-year-old groups, respectively. Among adolescents (participants 11- to 17-years old), most of the energy intake from liquids came from carbonated beverages, which accounted for an average of 207 kcal/day in this group (42% of their total energy intake from liquids). Health professionals should be attentive to the excessive consumption of sugar sweetened beverages in children and adolescents. The movement toward healthier dietary patterns at the individual and population levels may help to improve programs for preventing overweight and obesity in children and adolescents. Conclusion: From childhood to adolescence the daily volume of liquid ingested increased reaching a total of 2.0 liters on average. Of this volume, the daily volume of milk ingested decreased while the carbonated drinks, sweetened, nectars and artificial beverages increased significantly. The proportion of water remained constant in about 1/3 of the total volume. From 3 to 17 years of age the energy intake from carbonated beverages increased by about 20%. The carbonated drinks on average corresponded to a tenth of the daily requirements of energy of adolescents.

A semi-supervised classification technique based on interacting forces

Semi-supervised learning is a classification paradigm in which just a few labeled instances are available for the training process. To overcome this small amount of initial label information, the information provided by the unlabeled instances is also considered. In this paper, we propose a nature-inspired semi-supervised learning technique based on attraction forces. Instances are represented as points in a k-dimensional space, and the movement of data points is modeled as a dynamical system. As the system runs, data items with the same label cooperate with each other, and data items with different labels compete among them to attract unlabeled points by applying a specific force function. In this way, all unlabeled data items can be classified when the system reaches its stable state. Stability analysis for the proposed dynamical system is performed and some heuristics are proposed for parameter setting. Simulation results show that the proposed technique achieves good classification results on artificial data sets and is comparable to well-known semi-supervised techniques using benchmark data sets.

Determinação de vitaminas antioxidantes em suplementos e avaliação da rotulagem nutricional

Introdução: Diante das mudanças nos hábitos de consumo alimentar da população brasileira, suplementos vitamínicos e alimentos enriquecidos são veículos comumente empregados para atender as necessidades de ingestão de micronutrientes. A diversidade de suplementos vitamínicos comercializados atualmente leva à necessidade de desenvolvimento de métodos analíticos de fácil execução e alta produtividade. Informações confiáveis sobre os teores de vitaminas poderão ser obtidas somente com métodos analíticos validados. Objetivos: Validar metodologias analíticas e avaliar o teor de vitaminas antioxidantes em suplementos adquiridos no comércio do município de São Paulo - Brasil, o efeito do armazenamento nestes compostos e confrontar os valores analisados com os valores declarados na rotulagem. Métodos: As metodologias analíticas para determinação de vitaminas antioxidantes por cromatografia líquida de alta eficiência com detector de arranjo de diodos (CLAE-DAD) e de vitamina C por titulação potenciométrica foram validadas para as matrizes sólidas, oleosas e líquidas de suplementos vitamínicos. A estabilidade das vitaminas foi avaliada a cada 6 meses durante 12 meses de armazenamento e a avaliação da rotulagem foi realizada de acordo com as legislações vigentes no Brasil. Resultados: Para os métodos cromatográficos, os limites de detecção (LDs) e de quantificação (LQs) variaram entre 0,3 e 4,3 µg/mL, e entre 0,5 e 14,0 µg/mL respectivamente. As recuperações dos padrões adicionados nas matrizes variaram entre 92 por cento e 109 por cento e entre 86 por cento e 108 por cento no material de referência. A repetitividade foi calculada pelo desvio padrão relativo (RSD), apresentando valores entre 0,2 por cento e 9,6 por cento . Para a determinação de vitamina C pelo método potenciométrico, o LD e o LQ foram respectivamente 1 mg e 3 mg; a recuperação no material de referência foi de 99,8 por cento e a precisão variou entre 0,4 e 3,9 por cento . Das 57 amostras avaliadas, 59 por cento e 35 por cento apresentaram teores de vitaminas A e E respectivamente, abaixo dos valores declarados no rótulo; por outro lado, 20 por cento das amostras apresentaram teores de vitamina E acima dos valores declarados. Em relação aos teores de vitamina C, 60 por cento das amostras estavam de acordo com os valores declarados. O estudo da estabilidade demonstrou degradação significativa das vitaminas A, E e C em aproximadamente 90 por cento das amostras com 12 meses de armazenamento. Na avaliação da rotulagem dos suplementos vitamínicos, 47 das amostras apresentaram uma ou mais irregularidades. Conclusão: Os métodos propostos se mostraram adequados para análise de diferentes matrizes de suplementos vitamínicos. Os resultados das análises de vitaminas nestes produtos mostraram a necessidade urgente de monitoramento em conjunto com ações de fiscalização, pois verificou-se que a maioria das amostras não atenderam a legislação, principalmente quanto aos teores declarados na informação nutricional da rotulagem. A sobredosagem de vitaminas pode ser necessária para manter os teores declarados durante o armazenamento, porém, a quantidade adicional de vitamina a ser incluída no suplemento deve estar dentro de limites seguros e depende de cada amostra, pois além da matriz, diversos fatores relacionados aos compostos e à embalagem também podem influenciar na estabilidade das vitaminas.