Improving parameter learning of Bayesian nets from incomplete data

This paper addresses the estimation of parameters of a Bayesian network from incomplete data. The task is usually tackled by running the Expectation-Maximization (EM) algorithm several times in order to obtain a high log-likelihood estimate. We argue that choosing the maximum log-likelihood estimate (as well as the maximum penalized log-likelihood and the maximum a posteriori estimate) has severe drawbacks, being affected both by overfitting and model uncertainty. Two ideas are discussed to overcome these issues: a maximum entropy approach and a Bayesian model averaging approach. Both ideas can be easily applied on top of EM, while the entropy idea can be also implemented in a more sophisticated way, through a dedicated non-linear solver. A vast set of experiments shows that these ideas produce significantly better estimates and inferences than the traditional and widely used maximum (penalized) log-likelihood and maximum a posteriori estimates. In particular, if EM is adopted as optimization engine, the model averaging approach is the best performing one; its performance is matched by the entropy approach when implemented using the non-linear solver. The results suggest that the applicability of these ideas is immediate (they are easy to implement and to integrate in currently available inference engines) and that they constitute a better way to learn Bayesian network parameters.

Bycatch and discards of commercial trawl fisheries in the south coast of Portugal

Bycatch and discards are a cause of great concern in commercial world fisheries, with important ecological, economic and conservation implications. With the recent inclusion of a discards ban (‘landing obligation’), in the reform of the EU CFP, these issues have gained a tremendous attention from the economic, scientific, political and social point of view. Demersal trawl fisheries off the southern coast of Portugal capture an extraordinary diversity of species and generate considerable amounts of bycatch and discards. Bycatch includes commercially valuable target-species and bycatch species with low or no commercial value, but the great majority consists of unmarketable species, that are discarded. Bony fishes are dominant in bycatch and discards and the most discarded are of low or no commercial value. The reasons for discarding are fundamentally economic in nature (lack of commercial value) for bycatch species, and legal and administrative (legal minimum landing size) for commercially important species. The study of the reproductive biology of Galeus melastomus, discarded by crustacean trawls, suggests that a minimum landing size should be established for this species, and explains the importance of such a study in the assessment and management of fisheries. The discovery of a new species of the ray Neoraja iberica n. sp. contributes to the knowledge of the local marine biodiversity in Portuguese waters and of the global marine biodiversity. The three cases of abnormal hermaphroditism recorded in Etmopterus spinax, are the first cases known to date of hermaphroditism in this species. There is a need to find solutions to the problem of bycatch and discards of trawl fisheries in the Algarve coast. A combination of technical, regulatory and economic measures to minimize bycatch and reduce discards, before implementing a ‘landing obligation’, is thought to be the best approach to apply in the southern Portuguese multispecies trawl fisheries.

An integrated recycling approach for GFRP pultrusion wastes: recycling and reuse assessment into new composite materials using Fuzzy Boolean Nets

In this study, efforts were made in order to put forward an integrated recycling approach for the thermoset based glass fibre reinforced polymer (GPRP) rejects derived from the pultrusion manufacturing industry. Both the recycling process and the development of a new cost-effective end-use application for the recyclates were considered. For this purpose, i) among the several available recycling techniques for thermoset based composite materials, the most suitable one for the envisaged application was selected (mechanical recycling); and ii) an experimental work was carried out in order to assess the added-value of the obtained recyclates as aggregates and reinforcement replacements into concrete-polymer composite materials. Potential recycling solution was assessed by mechanical behaviour of resultant GFRP waste modified concrete-polymer composites with regard to unmodified materials. In the mix design process of the new GFRP waste based composite material, the recyclate content and size grade, and the effect of the incorporation of an adhesion promoter were considered as material factors and systematically tested between reasonable ranges. The optimization process of the modified formulations was supported by the Fuzzy Boolean Nets methodology, which allowed finding the best balance between material parameters that maximizes both flexural and compressive strengths of final composite. Comparing to related end-use applications of GFRP wastes in cementitious based concrete materials, the proposed solution overcome some of the problems found, namely the possible incompatibilities arisen from alkalis-silica reaction and the decrease in the mechanical properties due to high water-cement ratio required to achieve the desirable workability. Obtained results were very promising towards a global cost-effective waste management solution for GFRP industrial wastes and end-of-life products that will lead to a more sustainable composite materials industry.

Prescribed burning impact on forest soil properties – a fuzzy boolean nets approach

The Portuguese northern forests are often and severely affected by wildfires during the Summer season. These occurrences significantly affect and negatively impact all ecosystems, namely soil, fauna and flora. In order to reduce the occurrences of natural wildfires, some measures to control the availability of fuel mass are regularly implemented. Those preventive actions concern mainly prescribed burnings and vegetation pruning. This work reports on the impact of a prescribed burning on several forest soil properties, namely pH, soil moisture, organic matter content and iron content, by monitoring the soil self-recovery capabilities during a one year span. The experiments were carried out in soil cover over a natural site of Andaluzitic schist, in Gramelas, Caminha, Portugal, which was kept intact from prescribed burnings during a period of four years. Soil samples were collected from five plots at three different layers (0–3, 3–6 and 6–18) 1 day before prescribed fire and at regular intervals after the prescribed fire. This paper presents an approach where Fuzzy Boolean Nets (FBN) and Fuzzy reasoning are used to extract qualitative knowledge regarding the effect of prescribed fire burning on soil properties. FBN were chosen due to the scarcity on available quantitative data. The results showed that soil properties were affected by prescribed burning practice and were unable to recover their initial values after one year.

A qualitative description of soil parameters variation due to a prescribed fire in Portuguese northwestern forests using Fuzzy Boolean Nets — The case study of Cabreira mountain

In this paper we study the modifications that occurred in some forest soil properties after a prescribed fire. The research focused on the alterations of soil pH, soil moisture and soil organic matter content during a two-year span, from 2008 to 2009. The study site is located in Anjos, Vieira do Minho municipality, a forest site that has suffered from recurrent wildfires for several decades. Furze (Ulex, sp.), broom (Cytisus, sp.), gorse (Chamaespartum tridentatum) and a very few disperse adult pine (Pinus sylvestris) are the predominant vegetation type in the study area. The average height of this shrub vegetation is around 1.5 m. The prescribed fire was conducted by the National Forestry Authority (AFN) in November 2008. Fuzzy Boolean Nets (FBN) were used to evaluate the alteration in soil parameters when compared with adjacent spots where: i) no fire occurrence was registered since 1998; ii) fire occurrence was registered in 2008; and iii) vegetation pruning by mechanical cut was done in Spring six months prior to the prescribed fire event. Results suggest that in the particular case of the studied site, Anjos, the observed soil properties alterations cannot be related with the prescribed fire.

Fuzzy Boolean nets investigation of the alteration of some physical forest soil properties due to a prescribed fire action

Portuguese northern forests are often and severely affected by wildfires during the summer season. Some preventive actions, such as prescribed (or controlled) burnings and clear-cut logging, are often used as a measure to reduce the occurrences of wildfires. In the particular case of Serra da Cabreira forest, due to extremely difficulties in operational field work, the prescribed (or controlled) burning technique is the the most common preventive action used to reduce the existing fuel load amount. This paper focuses on a Fuzzy Boolean Nets analysis of the changes in some forest soil properties, namely pH, moisture and organic matter content, after a controlled fire, and on the difficulties found during the sampling process and how they were overcome. The monitoring process was conducted during a three-month period in Anjos, Vieira do Minho, Portugal, an area located in a contact zone between a two-mica coarse-grained porphyritic granite and a biotite with plagioclase granite. The sampling sites were located in a spot dominated by quartzphyllite with quartz veins whose bedrock is partially altered and covered by slightly thick humus, which maintains low undergrowth vegetation.

Temporal and spatial appearance of the membrane cytoskeleton and perineuronal nets in the rat neocortex.

Parvalbumin-immunoreactive interneurons are surrounded by perineuronal nets, containing molecules of the extracellular matrix (e.g. tenascin-R). Furthermore, they seem to have a special cytoskeleton composed of, among others, ankyrinR and beta Rspectrin. In the present developmental study we showed that the intracellular markers parvalbumin, ankyrinR and beta Rspectrin as well as Vicia Villosa agglutinin, an extracellular marker for perineuronal nets, appeared in the second postnatal week. In the third postnatal week, ankyrinR and beta R spectrin were present in the parvalbumin-positive interneurons. Tenascin-R appeared in a similar topographic distribution as the intracellular markers. The adult pattern was established upon the end of the fourth postnatal week. Our results indicate that cytoskeletal maturity maybe a prerequisite for the organization of perineuronal nets of extracellular matrix.

Flota de arrastre de fondo de menor escala en la Región Tumbes

Los problemas críticos de la pesquería de arrastre de menor escala, son las capturas de ejemplares juveniles, alta presencia de descartes, pesca incidental o accesoria y conflictos con los pescadores artesanales que usan redes de enmalle cortineras. En toda la zona de estudio, los índices de captura por unidad de esfuerzo (CPUE) fue 142,4 kg/h y 477,5 kg/lance, bycatch por unidad de esfuerzo (BPUE) fue 27,2 kg/h y 91,1 kg/lance. Los mayores CPUE fueron en la zona sur dentro de las 5 mn con 199,0 kg/h y 617,8 kg/lance. La composición de la captura relativa al peso estuvo dominada por el falso volador (Prionotus stephanophrys, 24,6%) y carajito (Diplectrum conceptione, 21,4%). Las especies incidentales más importantes fueron espejo (Selene peruviana, 9,8%), bereche (Larimus pacificus, 9,3%), cachema (Cynoscion analis, 4,0%), chiri (Peprilus medius, 2,9%), lenguado de boca chica (Etropus ectenes, 2,5%), doncella (Hemanthias peruanus, 2,1%). El descarte fue 19,1% de la captura, los principales recursos fueron merluza (Merluccius gayi peruanus, 39,1%), lengüeta (Symphurus sechurae, 10,9%), morena (Muraena clepsidra, 4,9%), pez hojita (Chloroscombrus orqueta, 4,8%), otras especies 31,5% (incluyendo restos de peces y equinodermos). El índice de impacto al ecosistema marino fue de 3,7 (1: no favorable al 10: favorable). Por lo que es un arte de pesca no amigable con el ecosistema marino que no debe usarse dentro del área costera

Recognition of Facial Expressions with Autoencoders and Convolutional-Nets

Les humains communiquent via différents types de canaux: les mots, la voix, les gestes du corps, des émotions, etc. Pour cette raison, un ordinateur doit percevoir ces divers canaux de communication pour pouvoir interagir intelligemment avec les humains, par exemple en faisant usage de microphones et de webcams. Dans cette thèse, nous nous intéressons à déterminer les émotions humaines à partir d’images ou de vidéo de visages afin d’ensuite utiliser ces informations dans différents domaines d’applications. Ce mémoire débute par une brève introduction à l'apprentissage machine en s’attardant aux modèles et algorithmes que nous avons utilisés tels que les perceptrons multicouches, réseaux de neurones à convolution et autoencodeurs. Elle présente ensuite les résultats de l'application de ces modèles sur plusieurs ensembles de données d'expressions et émotions faciales. Nous nous concentrons sur l'étude des différents types d’autoencodeurs (autoencodeur débruitant, autoencodeur contractant, etc) afin de révéler certaines de leurs limitations, comme la possibilité d'obtenir de la coadaptation entre les filtres ou encore d’obtenir une courbe spectrale trop lisse, et étudions de nouvelles idées pour répondre à ces problèmes. Nous proposons également une nouvelle approche pour surmonter une limite des autoencodeurs traditionnellement entrainés de façon purement non-supervisée, c'est-à-dire sans utiliser aucune connaissance de la tâche que nous voulons finalement résoudre (comme la prévision des étiquettes de classe) en développant un nouveau critère d'apprentissage semi-supervisé qui exploite un faible nombre de données étiquetées en combinaison avec une grande quantité de données non-étiquetées afin d'apprendre une représentation adaptée à la tâche de classification, et d'obtenir une meilleure performance de classification. Finalement, nous décrivons le fonctionnement général de notre système de détection d'émotions et proposons de nouvelles idées pouvant mener à de futurs travaux.

Studies on the exploratory trawl fishing and significance of selectivity in trawl fishing gear for conservation of marine fishery resources with special reference to the inshore waters off Mangalore

School of Industrial Fisheries, Cochin University of Science and Technology

Gill Nets of Kerala

The fishing industry the world over is passing through a critical situation.The landings of marine fishes seemed to have reached saturation in major fishing areas of the world.As a general rule fisheries employing fishing gear and techniques used by small scale and artisanal fishermen either from shore or from onboard small fishing craft come under small-scale fisheries.This study on gill nets of Kerala, the fishing method depended upon by maximum fishermen of the state focuses on the importance of this selective and low energy fishing method in the marine fishing sector of the state.The study opens with the conceptual framework by briefly reviewing the crisis in the marine fisheries sector. Maximum fishermen depend upon gill net, which is, an important selective and low energy fishing gear. A review of relevant literature on aspects such as material, selectivity and techno-economic efficiency together with scope and main objectives of the study form the major part of the compass of the introductory chapter.This survey provided the inputs for selection of centres. The chapter presents the basis for selection of sample centres, sample units and methodology for field and experimental study.The subject matter of the fourth chapter is a basic study on gear aterials. The weathering resistance, which is an important criterion to assess the material performance, was studied for polyamide monofilament in comparison to polyamide multifilament and polyethylene twisted monofilament.The study provides supporting evidence of oxidation and characteristic C-O stretching in polyethylene and cyclic lactam .formation and presence of OH in polyamide.The study indicates that small mesh gill netting can be encouraged as a selective fishing method in the inshore waters with restrained use of 30 and 32 mm mesh sizes. The economic efficiency was assessed using standard indices such as rate of return, internal rate of return, pay back period, fishery income, energy efficiency and factor productivity. The effect of size and cost of capital and cost of production on the economics of operation is also discussed in this chapter. It was observed that level of technology did not have direct effect on economic performance.

Studies on the technological aspects and economic efficiency of trawl gear operating along the Cochin coast

Trawling, despite being heavily energy expensive, still continues to be the most energy expensive fishing method particularly so in View of the export oriented nature of the Indian seafood industry. This study therefore aims at analyzing the efficiency of trawls operation from Cochin, an important fishing center along the southwest coast of India. The analysis is made along two perspectives - economic and technological. Even though technological efficiency complement economic efficiency, in the fishing parlance, parameters like the size composition of the catch, selectivity factors, etc., will have a direct bearing on the technological qualities of the trawl, and which parameters will have a significant impact on the effective exploitation of a fishery stock. Whereas the technological analysis aims at improving the efficiency with regard to the effective utilization of fuel and fishery stocks, economic analysis ascertains the present status of the trawling operations from the commercial angle.

studies on Prawn Gill Nets of The Kerala Coast

In the present study. extensive investigations were carried out on various factors affecting the selectivity of prawn gill nets with reference to material, mesh size, coefficient of hanging secolouration. Effect of tidal current on fishing height of prawn gill net and seasonal variation of catch during the course of these investigations were also studied.

Materials for Fish Nets-Their Properties, Selection and Preservation

The properties of synthetic fibres vary with thc inherent physical characteristics of the basic raw materials used mode of preparation of yarns and method of construction of twines. Since the synthetic fibres as maufactured from polymers which are synthesized from simple chemical units, the qualities of man-made fibres can he influenced by the process of manufacture and certain modifications can even be introduced at the processing stage to meet any specific requirement to a certain extent. Hence, an elaborate study of the properties of fish not twines produced has been taken up with a view to determining their suitability for various types of fishing gear with particular reference to conditions prevailing in India.