Particle Competition and Cooperation in Networks for Semi-Supervised Learning

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

Particle Competition and Cooperation in Networks for Semi-Supervised Learning with Concept Drift

Concept drift is a problem of increasing importance in machine learning and data mining. Data sets under analysis are no longer only static databases, but also data streams in which concepts and data distributions may not be stable over time. However, most learning algorithms produced so far are based on the assumption that data comes from a fixed distribution, so they are not suitable to handle concept drifts. Moreover, some concept drifts applications requires fast response, which means an algorithm must always be (re) trained with the latest available data. But the process of labeling data is usually expensive and/or time consuming when compared to unlabeled data acquisition, thus only a small fraction of the incoming data may be effectively labeled. Semi-supervised learning methods may help in this scenario, as they use both labeled and unlabeled data in the training process. However, most of them are also based on the assumption that the data is static. Therefore, semi-supervised learning with concept drifts is still an open challenge in machine learning. Recently, a particle competition and cooperation approach was used to realize graph-based semi-supervised learning from static data. In this paper, we extend that approach to handle data streams and concept drift. The result is a passive algorithm using a single classifier, which naturally adapts to concept changes, without any explicit drift detection mechanism. Its built-in mechanisms provide a natural way of learning from new data, gradually forgetting older knowledge as older labeled data items became less influent on the classification of newer data items. Some computer simulation are presented, showing the effectiveness of the proposed method.

Particle competition and cooperation to prevent error propagation from mislabeled data in semi-supervised learning

Semi-supervised learning is applied to classification problems where only a small portion of the data items is labeled. In these cases, the reliability of the labels is a crucial factor, because mislabeled items may propagate wrong labels to a large portion or even the entire data set. This paper aims to address this problem by presenting a graph-based (network-based) semi-supervised learning method, specifically designed to handle data sets with mislabeled samples. The method uses teams of walking particles, with competitive and cooperative behavior, for label propagation in the network constructed from the input data set. The proposed model is nature-inspired and it incorporates some features to make it robust to a considerable amount of mislabeled data items. Computer simulations show the performance of the method in the presence of different percentage of mislabeled data, in networks of different sizes and average node degree. Importantly, these simulations reveals the existence of the critical points of the mislabeled subset size, below which the network is free of wrong label contamination, but above which the mislabeled samples start to propagate their labels to the rest of the network. Moreover, numerical comparisons have been made among the proposed method and other representative graph-based semi-supervised learning methods using both artificial and real-world data sets. Interestingly, the proposed method has increasing better performance than the others as the percentage of mislabeled samples is getting larger. © 2012 IEEE.

Fuzzy community structure detection by particle competition and cooperation

Identification and classification of overlapping nodes in networks are important topics in data mining. In this paper, a network-based (graph-based) semi-supervised learning method is proposed. It is based on competition and cooperation among walking particles in a network to uncover overlapping nodes by generating continuous-valued outputs (soft labels), corresponding to the levels of membership from the nodes to each of the communities. Moreover, the proposed method can be applied to detect overlapping data items in a data set of general form, such as a vector-based data set, once it is transformed to a network. Usually, label propagation involves risks of error amplification. In order to avoid this problem, the proposed method offers a mechanism to identify outliers among the labeled data items, and consequently prevents error propagation from such outliers. Computer simulations carried out for synthetic and real-world data sets provide a numeric quantification of the performance of the method. © 2012 Springer-Verlag.

Particle competition and cooperation for semi-supervised learning with label noise

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

Semi-supervised learning with concept drift using particle dynamics applied to network intrusion detection data

Concept drift, which refers to non stationary learning problems over time, has increasing importance in machine learning and data mining. Many concept drift applications require fast response, which means an algorithm must always be (re)trained with the latest available data. But the process of data labeling is usually expensive and/or time consuming when compared to acquisition of unlabeled data, thus usually only a small fraction of the incoming data may be effectively labeled. Semi-supervised learning methods may help in this scenario, as they use both labeled and unlabeled data in the training process. However, most of them are based on assumptions that the data is static. Therefore, semi-supervised learning with concept drifts is still an open challenging task in machine learning. Recently, a particle competition and cooperation approach has been developed to realize graph-based semi-supervised learning from static data. We have extend that approach to handle data streams and concept drift. The result is a passive algorithm which uses a single classifier approach, naturally adapted to concept changes without any explicit drift detection mechanism. It has built-in mechanisms that provide a natural way of learning from new data, gradually "forgetting" older knowledge as older data items are no longer useful for the classification of newer data items. The proposed algorithm is applied to the KDD Cup 1999 Data of network intrusion, showing its effectiveness.

Effects of corn particle size and physical form of the diet on the gastrointestinal structures of broiler chickens

The objective of this experiment was to investigate the effects of different particle sizes, expressed as Geometric Mean Diameter (GMD) of corn (0.336mm, 0.585mm, 0.856mm and 1.12mm) of mash and pelleted broiler chicken diets on the weight of the gizzard, duodenum and jejunum+ileum; on the pH of the gizzard and small intestine and on the characteristics of the duodenal mucous layer (number and height of villi and crypt depth) in 42-day-old broilers. The physical form and the particle size of the diet had no significant effect on gizzard and intestine pH (p > 0.05). A greater gizzard weight was seen in the birds receiving pelleted diet and particle size of 0.336mm (p < 0.008). However, for the particle sizes of 0.856 and 1.12 mm, a greater weight was found in birds that received mash diet (p < 0.039 and p < 0.006, respectively). Also, gizzard weight was greater with increasing corn GMD independent of the physical form of the diet. In the mash diet, the increase in particle size promoted a quadratic response in the weight of duodenum and jejunum + ileum. The pelleted diet promoted a greater number of villi per transverse duodenum cut (p < 0.007) and greater crypt depth (p < 0.05). As the particle size increased, there was a linear increase of villus height and crypt depth in the duodenum, irrespective of the physical form of the diet.

Photoluminescence of SrTiO3: Influence of Particle Size and Morphology

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

Effects of the particle size and nucleation temperature on tellurite 20Li(2)O-80TeO(2) glass crystallization

20Li(2)O-80TeMO(2) glasses were heat annealed at different temperatures between T-g and T-x and studied by using XRD, FTIR spectroscopy and DSC techniques to understand the crystallization kinetics in this glass matrix. The infrared band structure of this glass is similar to what was observed in glassy TeO2. XRD results reveal the presence of three distinct crystalline gamma-TeO2, alpha-TeO2 and Li2Te2O5 phases during the crystallization process. This is a first report of gamma-TeO2 phase crystallization in this glass matrix. DSC results confinn the crystallization of three distinct structures in the glass. In summary, our results suggest a crystallization hierarchy on this glass matrix since the gamma-TeO2 and alpha-TeO2 phases crystallization occurs before the Li2Te2O5 phase crystallization. (c) 2006 Elsevier B.V. All rights reserved.

Competition and phylogeny determine community structure in Mullerian co-mimics

Until recently, the study of negative and antagonistic interactions (for example, competition and predation) has dominated our understanding of community structure, maintenance and assembly(1). Nevertheless, a recent theoretical model suggests that positive interactions (for example, mutualisms) may counterbalance competition, facilitating long-term coexistence even among ecologically undifferentiated species(2). Mullerian mimics are mutualists that share the costs of predator education(3) and are therefore ideally suited for the investigation of positive and negative interactions in community dynamics. The sole empirical test of this model in a Mullerian mimetic community supports the prediction that positive interactions outweigh the negative effects of spatial overlap(4) (without quantifying resource acquisition). Understanding the role of trophic niche partitioning in facilitating the evolution and stability of Mullerian mimetic communities is now of critical importance, but has yet to be formally investigated. Here we show that resource partitioning and phylogeny determine community structure and outweigh the positive effects of Mullerian mimicry in a species-rich group of neotropical catfishes. From multiple, independent reproductively isolated allopatric communities displaying convergently evolved colour patterns, 92% consist of species that do not compete for resources. Significant differences in phylogenetically conserved traits (snout morphology and body size) were consistently linked to trait-specific resource acquisition. Thus, we report the first evidence, to our knowledge, that competition for trophic resources and phylogeny are pivotal factors in the stable evolution of Mullerian mimicry rings. More generally, our work demonstrates that competition for resources is likely to have a dominant role in the structuring of communities that are simultaneously subject to the effects of both positive and negative interactions.

Effect of Airborne-Particle Abrasion and Mechanico-Thermal Cycling on the Flexural Strength of Glass Ceramic Fused to Gold or Cobalt-Chromium Alloy

Purpose: To evaluate the effect of airborne-particle abrasion and mechanico-thermal cycling on the flexural strength of a ceramic fused to cobalt-chromium alloy or gold alloy.Materials and Methods: Metallic bars (n = 120) were made (25 mm x 3 mm x 0.5 mm): 60 with gold alloy and 60 with Co-Cr. At the central area of the bars (8 mm x 3 mm), a layer of opaque ceramic and then two layers of glass ceramic (Vita VM13, Vita Zahnfabrick) were fired onto it (thickness: 1 mm). Ten specimens from each alloy group were randomly allocated to a surface treatment [(tungsten bur or air-particle abrasion (APA) with Al(2)O(3) at 10 mm or 20 mm away)] and mechanico-thermal cycling (no cycling or mechanically loaded 20,000 cycles; 10 N distilled water at 37 degrees C and then thermocycled 3000 cycles; 5 degrees C to 55 degrees C, dwell time 30 seconds) combination. Those specimens that did not undergo mechanico-thermal cyclingwere stored inwater (37 degrees C) for 24 hours. Bond strength was measured using a three-point bend test, according to ISO 9693. After the flexural strength test, failure types were noted. The data were analyzed using three factor-ANOVA and Tukey's test (alpha = 0.05).Results: There were no significant differences between the flexural bond strength of gold and Co-Cr groups (42.64 +/- 8.25 and 43.39 +/- 10.89 MPa, respectively). APA 10 and 20 mm away surface treatment (45.86 +/- 9.31 and 46.38 +/- 8.89 MPa, respectively) had similar mean flexural strength values, and both had significantly higher bond strength than tungsten bur treatment (36.81 +/- 7.60 MPa). Mechanico-thermal cycling decreased the mean flexural strength values significantly for all six alloy-surface treatment combinations tested when compared to the control groups. The failure type was adhesive in the metal/ceramic interface for specimens surface treated only with the tungsten bur, and mixed for specimens surface treated with APA 10 and 20 mm.Conclusions: Considering the levels adopted in this study, the alloy did not affect the bond strength; APA with Al(2)O(3) at 10 and 20 mm improved the flexural bond strength between ceramics and alloys used, and the mechanico-thermal cycling of metal-ceramic specimens resulted in a decrease of bond strength.

Forward energy flow, central charged-particle multiplicities, and pseudorapidity gaps in W and Z boson events from pp collisions at root s=7 TeV

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Finns growth dynamics, competition and power-law scaling

We study the growth dynamics of the size of manufacturing firms considering competition and normal distribution of competency. We start with the fact that all components of the system struggle with each other for growth as happened in real competitive business world. The detailed quantitative agreement of the theory with empirical results of firms growth based on a large economic database spanning over 20 years is good with a single set of the parameters for all the curves. Further, the empirical data of the variation of the standard deviation of the growth rate with the size of the firm are in accordance with the present theory rather than a simple power law. (C) 2003 Elsevier B.V. B.V. All rights reserved.

A new method to control particle size and particle size distribution of SnO2 nanoparticles for gas sensor applications

The controlled growth of SnO2 nanoparticles for gas sensor applications is reported by these authors. Nb2O5 additive is used to control nucleation and growth of the SnO2 (see Figure), which is synthesized by the polymeric precursor method. Preliminary gas sensing measurements are performed and it is demonstrated that the response time of the Nb2O5-doped SnO2 is faster than that of the undoped material.

Interference competition and scavenging by Crematogaster ants (Hymenoptera: Formicidae) associated with the webs of the social spider Anelosimus eximius (Araneae: Theridiidae) in the central Amazon

Crematogaster cf. victima is a common inhabitant in the sheet web nests of the social spider Anelosimus eximius in the central Amazon basin near Manaus. A number of other ant species were found foraging on the non-sticky webs of A. eximius, but none of these reached the web occupation frequency found in C. cf. victima, nor, with the exception of an unidentified species of Pheidole, did they form satellite nests in the web, as did this species. Many prey which escaped the knock-down threads of the sheet web of A. eximius colonies were captured by ants in the lower web portions which they dominated. Furthermore, prey which were rejected by A. eximius, especially large, heavily sclerotized beetles, were also consumed by this ant. Repeated observations and experiments suggest that C. cf. victima is able to deter A. eximius activity through aerial venom release. Resources lost by A. eximius colonies to ants, especially C. cf. victima, in colonial web area and prey, may pose significant costs and may reduce colony growth.