Search

Context: Classification of software requirements into different categories is a critically important task in requirements engineering (RE). Developing machine learning (ML) approaches for requirements classification has attracted great interest in the RE community since the 2000s. Objective: This paper aims to address two related problems that have been challenging real-world applications of ML approaches: the problems of class imbalance and high dimensionality with low sample size data (HDLSS). These problems can greatly degrade the classification performance of ML methods. Method: The paper proposes HC4RC, a novel ML approach for multiclass classification of requirements. HC4RC solves the aforementioned problems through semantic-role-based feature selection, dataset decomposition and hierarchical classification. We experimentally compare the effectiveness of HC4RC with three closely related approaches - two of which are based on a traditional statistical classification model whereas one uses an advanced deep learning model. Results: Our experiment shows: 1) The class imbalance and HDLSS problems present a challenge to both traditional and advanced ML approaches. 2) The HC4RC approach is simple to use and can effectively address the class imbalance and HDLSS problems compared to similar approaches. Conclusion: This paper makes an important practical contribution to addressing the class imbalance and HDLSS problems in multiclass classification of software requirements., 36 pages, 3 tables, 4 figures

Public policy for the preservation and development of traditional villages in China has witnessed a shift. That is from the equal distribution of finances to officially recognised traditional villages to the prioritisation of the development of more potential traditional villages that stimulate the development of surrounding villages and regions. This study takes 206 traditional villages in Hubei Province as an example and proposes a method framework for ‘Feature identification, Hierarchical classification (FIHC) that combines field research, spatial geography and spatial network methods. The results of spatial geography show that traditional villages in Hubei province are unevenly distributed, mainly along the mountain ranges in the east and west, with scattered distribution in the central plains. The kernel density result reveals “3 + 2 high-density hotspot zones” (A, B, and C are high-density areas, whereas D and E are low-density areas). Furthermore, the results of the spatial network appear that the traditional villages’ ecological situation in Zone A is better than in other areas, but their transportation accessibility and economy are poor; Zone B villages’ transportation accessibility is better than in other areas; Zone C has the best economy; and three conditions of traditional villages in the D and E zones are poor. FIHC can identify villages with more development potential and stratify these villages in a multidimensional way. It is innovative and an important contribution to policymakers and planners in developing ‘phased and focused’ public policies and rural planning.

In the taxonomic trace links concept, the source and target artifacts are connected through knowledge organization structure (e.g., taxonomy). We introduce in this paper a recommender system that recommends labels to requirements artifacts from domain-specific taxonomy to establish taxonomic trace links. The tool exploits the hierarchical nature of taxonomies and uses requirements text and context information as input to the recommender. © 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).

Using hierarchical clustering with unweighted pair-group method with arithmetic mean (UPGMA) we arranged 602 phytosociological releves of beech forests on the present forest line, mainly from the Julian Alps and the Trnovo Forest Plateau (we also included the releves from the Karawanks and the Kamnik Alps) into 32 clusters. Based on their analysis and comparison with previously described similar (alti)montane-subalpine beech communities we classified most of the releves into the association Polysticho lonchitis-Fagetum and its new subassociations ericetosum carneae, cardaminetosum trifoliae, luzuletosum niveae, luzuletosum luzuloidis, calamagrostieteosum variae, allietosum victorialis, adoxetosum moschatellinae, stellarietosum nemorum and several new variants. The altitude of the studied stands is predominantly 1400 to 1550 m (the upper line is at 1660 m); they occur at all aspects, frequently on steep and very steep slopes, mainly on limestone and dolomite limestone, the predominant soil type is rendzina. These stands are species rich (on average 61 species per releve, altogether more than 500 vascular plants) and have many species in common with the stands of associations Rhododendro hirsuti-Fagetum and Rhodothamno-Laricetum. They belong to the habitat type of Community interest 91K0 Illyrian Fagus sylvatica forests and are some of the best preserved forest communities in the Southeastern Alps and northwestern Dinaric Alps, comprising small areas of virgin forest remnants and performing an important protective role as a biotope.

Line-Line (LL) and Line-Ground (LG) faults may not be detected by common protection devices in PV arrays due to these faults are not detectable under high fault impedance and low mismatch level. In recent years, many efforts have been devoted to overcome these challenges using intelligent methods. However, these methods could not classify the type of faults and diagnose their severity. This paper proposes a novel and intelligent fault monitoring method to detect and classify LL and LG faults at the DC side of PV systems. For this purpose, the main features of Current-Voltage (I-V) curves under different fault events and normal conditions are extracted. The faults are categorized using the Hierarchical Classification (HC) platform. Later, the LL and LG faults are detected and classified by Machine Learning (ML) methods. The proposed method aims to reduce the amount of dataset which is required for the learning process and also obtain a higher accuracy in detecting and classifying the fault events at low mismatch levels and high fault impedance compared to other fault diagnostic methods. The experimental results verify that the proposed method precisely detects and classifies LL and LG faults on PV systems under the different conditions and severity with the accuracy of 96.66% and 91.66%, respectively.

Automatic identification of plant specimens from amateur photographs could improve species range maps, thus supporting ecosystems research as well as conservation efforts. However, classifying plant specimens based on image data alone is challenging: some species exhibit large variations in visual appearance, while at the same time different species are often visually similar; additionally, species observations follow a highly imbalanced, long-tailed distribution due to differences in abundance as well as observer biases. On the other hand, most species observations are accompanied by side information about the spatial, temporal and ecological context. Moreover, biological species are not an unordered list of classes but embedded in a hierarchical taxonomic structure. We propose a multimodal deep learning model that takes into account these additional cues in a unified framework. Our Digital Taxonomist is able to identify plant species in photographs better than a classifier trained on the image content alone, the performance gained is over 6 percent points in terms of accuracy., ISPRS Journal of Photogrammetry and Remote Sensing, 182, ISSN:0924-2716

Multi-label classification assigns multiple labels to each document concurrently. Many real-world classification problems tend to employ high-dimensional label spaces, which can be naturally structured in a hierarchy. In this type of problem, each instance may belong to multiple labels and labels are organized in a hierarchical structure. It presents a more complex problem than flat classification, given that the classification algorithm has to take into account hierarchical relationships between labels and be able to predict multiple labels for the same instance. Few studies have investigated multi-label text classification for the Arabic language. Most of these studies have focused mainly on flat classification and have neglected the hierarchical structure. Therefore, this paper explores the hierarchical multi-label classification in the context of the Arabic language. It proposes a hierarchical multi-label Arabic text classification (HMATC) model with a machine learning approach. The impact of feature selection methods and feature set dimensions on classification performance are also investigated. In addition, the Hierarchy Of Multilabel ClassifiER (HOMER) algorithm is optimized via examination of different sets of multi-label classifiers, clustering algorithms and different numbers of clusters to improve the hierarchical classification. Moreover, this study contributes to existing research by introducing a hierarchical multi-label Arabic dataset in an appropriate format for hierarchical classification and making it publicly available. The results reveal that the proposed model outperforms all models considered in the experiments in terms of the computational cost, which consumed less cost (2 h) compared with other evaluated models. In addition, it shows a significant improvement compared with the state-of-the-art model (Fatwa model) in terms of Hamming loss (0.004), hierarchical loss (1.723), multi-label accuracy (0.758), subset accuracy (0.292), micro-averaged precision (0.879), micro-averaged recall (0.828), and micro-averaged F-measure (0.853).

International audience; Crop supply and management is a global issue, particularly in the context of global climate change and rising urbanization. Accurate mapping and monitoring of specific crop types are crucial for crop studies. In this study, we proposed: (1) a methodology to map two main winter crops (winter wheat and winter barley) in the northern region of Finistère with high-resolution Sentinel-2 data. Different classification approaches (the hierarchical classification and the classical direct extraction), and classification methods (pixel-based classification (PBC) and object-based classification (OBC)) were performed and evaluated. Subsequently, (2) a further study that involved monitoring the phenology of the winter crops was carried out, based on the previous results. The aim is to understand the temporal behavior from sowing to harvesting, identifying three important phenological statuses (germination, heading, and ripening, including harvesting). Due to the high frequency of precipitation in our study area, crop phenology monitoring was performed using Sentinel-1 C-band SAR backscatter time series data using the Google Earth Engine (GEE) platform. The results of the classification showed that the hierarchical classification achieved a better accuracy when it is compared to the direct extraction, with an overall accuracy of 0.932 and a kappa coefficient of 0.888. Moreover, in the hierarchical classification process, OBC reached a better accuracy in cropland mapping, and PBC was proven more suitable for winter crop extraction. Additionally, in the time series backscatter coefficient of winter wheat, the germination and ripening (harvesting) phases can be identified at VV and VH/VV polarizations, and heading can be identified in both VV and VH polarizations. Secondly, we were able to detect the germination phase of winter barley in VV and VH, ripening with both polarizations and VH/VV, and finally, heading in VV and VH polarizations.

Liver ultrasound (US) plays a critical role in diagnosing liver diseases. However, it is often difficult for examiners to accurately identify the liver segments captured in US images due to patient variability and the complexity of US images. Our study aim is automatic, real-time recognition of standardized US scans coordinated with reference liver segments to guide examiners. We propose a novel deep hierarchical architecture for classifying liver US images into 11 standardized US scans, which has yet to be properly established due to excessive variability and complexity. We address this problem based on a hierarchical classification of 11 US scans with different features applied to individual hierarchies as well as a novel feature space proximity analysis for handling ambiguous US images. Experiments were performed using US image datasets obtained from a hospital setting. To evaluate the performance under patient variability, we separated the training and testing datasets into distinct patient groups. The experimental results show that the proposed method achieved an F1-score of more than 93%, which is more than sufficient for a tool to guide examiners. The superior performance of the proposed hierarchical architecture was demonstrated by comparing its performance with that of non-hierarchical architecture.

In this paper, we propose a hierarchical multi-modal multi-label attribute classification model for anime illustrations using a graph convolutional network (GCN). Our focus is on the challenging task of multi-label attribute classification, which requires capturing subtle features intentionally highlighted by creators of anime illustrations. To address the hierarchical nature of these attributes, we leverage hierarchical clustering and hierarchical label assignments to organize the attribute information into a hierarchical feature. The proposed GCN-based model effectively utilizes this hierarchical feature to achieve high accuracy in multi-label attribute classification. The contributions of the proposed method are as follows. Firstly, we introduce GCN to the multi-label attribute classification task of anime illustrations, enabling the capturing of more comprehensive relationships between attributes from their co-occurrence. Secondly, we capture subordinate relationships among the attributes by adopting hierarchical clustering and hierarchical label assignment. Lastly, we construct a hierarchical structure of attributes that appear more frequently in anime illustrations based on certain rules derived from previous studies, which helps to reflect the relationships between different attributes. The experimental results on multiple datasets show that the proposed method is effective and extensible by comparing it with some existing methods, including the state-of-the-art method.

Over the past decades, the understanding of peptides and proteins biological functions has been an active research topic. Latest research works in this field have suggested that protein conformations may be a key feature for gaining insights into protein biological functions. However, analyzing small and highly flexible protein chunks, namely oligopeptides made of a handful of amino acids, remains challenging because of their dynamics and wide range of conformations. In this paper, a statistical methodology based on unsupervised statistical learning is proposed for analyzing 3D conformations small and highly flexible elastin-derived peptides. The goal of this study is twofold: first, is it aimed at identifying the most frequent conformations of each peptide and to study their stability. Second, and most important, it is aimed at comparing main conformations of different elastin-derived peptides to identify the “signature” than can be linked to a biological activity. The main strength of the present work is to propose a method for confirmation recognition that is not affected by peptide rotations or translations and, hence, avoids the use of the complex superposition methods. In addition, the proposed approach uses Kernel PCA to eliminate atypical peptide conformations. Due to the instability of those peptides, removing outliers is crucial since they may dramatically impact clustering results. To extract the most frequent conformations, we propose to use a hierarchical clustering method. Eventually, a peptide activity detector is defined based on comparison of main conformation found in different peptides. The main interests of the proposed method are twofold: first, it is fully automatic method, second, it does not require any additional information or expertise and, third, it can identify conformations accurately that make peptides enabling a given biological activity. Experimental results on a large dataset of peptides conformations highlight the relevance and efficiency of the proposed method.

(1) Background: in recent years, a lot of the research of statistical methods focused on the classification problem in presence of imprecise data. A particular case of imprecise data is the interval-valued data. Following this research line, in this work a new hierarchical classification technique for multivariate interval-valued data is suggested for diagnosis of the breast cancer; (2) Methods: an unsupervised hierarchical classification method for imprecise multivariate data (called HC-ID) is performed for diagnosis of breast cancer (i.e., to discriminate between benign or malignant masses) and the results have been compared with the conventional (unsupervised) hierarchical classification approach (HC); (3) Results: the application on real data shows that the HC-ID procedure performs better HC procedure in terms of accuracy (HC-ID = 0.80, HC = 0.66) and sensitivity (HC-ID = 0.61, HC = 0.08). In the results obtained by the usual procedure, there is a high degree of false-negative (i.e., benign cancer diagnosis in malignant status) affected by the high degree of variability (i.e., uncertainty) characterizing the worst data.

The Portuguese Navy gave financial support to THEMIS project under the aim of the development of a decision support system to get optimal decisions in short time in a disaster context, optimizing the decision chain, allowing to get a better performance of tasks execution allowing a reduction of costs. In [14, 17], the authors have considered the facilities and high qualified staff of Portuguese Navy and proposed a variant of the Delphi method, a method that is exceptionally useful where the judgments of individuals are considered as an important information source. They proposed a system that prioritize certain teams for specific incidents taking into account the importance of each team that acts in case of emergency. In the present work we propose a distinct method of computing the weights that represent the importance given to experts opinion in the Delphi method used in [14, 17] under the idea that shall not depend on the years of experience of each expert exclusively but also shall be considered the kind of expert experience. To justify this suggestion we have used hierarchical classification, allowing to identify different padrons for experts with the “same experience”. Also discriminant analysis and multidimensional scaling revealed to be adequate techniques for this issue. We can classify the experience of each expert evaluating the similarity/distance between the individuals in the group of proposed experts and compare with the number of consensus presented. In this manuscript we propose an alternative way of weighting the experts experience that contributes to a decision support system capable to prioritize a set of teams for certain disaster incidents involving maritime issues. The decision support system is still been tested but, with this work, we hope to have given an improvement to its optimization.

The role of classification in science and other spheres can hardly be underestimated. Accounting is no exception. The requirements to “natural classification” were formulated in the late twentieth century and can be used for further development and improvement of accounting. The article is devoted to the analysis of the accounting system for compliance with the requirements of natural classification. Accounting information is generated on the basis of classification, with accounts themselves used as classification criteria. The existing system of accounting was analyzed for compliance with the major twelve natural classification requirements to find out that the system is totally incompliant with five of these requirements and five more are met only partially. In order to analyze the accounting system for compliance with the remaining two requirements it is necessary to start with decomposition of accounts into separate classification groups. The methods of scientific classification, formal logic, comparison, analysis and synthesis are used in this work. The results of the study can be used to improve the organization and planning of scientific research in the field of accounting. The research is of practical significance as it provides the recommendations on changing the composition and special features of accounts to meet the requirements of natural classification which in turn will reveal the very essence of the subject area in multidimensional and logical way.

Speaker identification refers to the process of recognizing human voice using artificial intelligence techniques. Speaker identification technologies are widely applied in voice authentication, security and surveillance, electronic voice eavesdropping, and identity verification. In the speaker identification process, extracting discriminative and salient features from speaker utterances is an important task to accurately identify speakers. Various features for speaker identification have been recently proposed by researchers. Most studies on speaker identification have utilized short-time features, such as perceptual linear predictive (PLP) coefficients and Mel frequency cepstral coefficients (MFCC), due to their capability to capture the repetitive nature and efficiency of signals. Various studies have shown the effectiveness of MFCC features in correctly identifying speakers. However, the performances of these features degrade on complex speech datasets, and therefore, these features fail to accurately identify speaker characteristics. To address this problem, this study proposes a novel fusion of MFCC and time-based features (MFCCT), which combines the effectiveness of MFCC and time-domain features to improve the accuracy of text-independent speaker identification (SI) systems. The extracted MFCCT features were fed as input to a deep neural network (DNN) to construct the speaker identification model. Results showed that the proposed MFCCT features coupled with DNN outperformed existing baseline MFCC and time-domain features on the LibriSpeech dataset. In addition, DNN obtained better classification results compared with five machine learning algorithms that were recently utilized in speaker recognition. Moreover, this study evaluated the effectiveness of one-level and two-level classification methods for speaker identification. The experimental results showed that two-level classification presented better results than one-level classification. The proposed features and classification model for identifying a speaker can be widely applied to different types of speaker datasets.

Due to different combinations of agriculture, livestock and forestry managed by rotation, succession and intercropping practices, integrated agriculture production systems such as integrated crop–livestock systems (iCL) constitute a very complex target and a challenge for automatic mapping of cropping practices based on remote sensing data. The overall objective of this study was to develop a classification strategy for the annual mapping of integrated Crop–Livestock systems (iCL) at a regional scale. This strategy was designed and tested in the six agro-climatic regions of Mato Grosso, the largest Brazilian soybean producer state, using MODIS satellite time-series images acquired between 2012 and 2019, ground data with heterogeneous distribution in space and time and a Random Forest classifier. The results showed that: 1. the use of unbalanced training samples with a class composition close to the real one was the right classifier training strategy; 2. the use of a single training database (pooling samples from different years and regions) to classify each region and year individually proved to be robust enough to provide similar classification accuracies in comparison to those based on the use of a database acquired for each region and for each year. The final hierarchical classification overall accuracy was 0.89 for Level 1, the cropping pattern level (single and double crops DC); 0.84 for Level 2, the DC category level (integrated system iCL soy-pasture/brachiaria, soy-cotton and soy-cereal); 0.77 for Level 3, the iCL level (iCL1 soy-pasture and iCL2 soy-pasture mixed with corn). The F-scores for DC, iCL and iCL1 cropping systems presented high accuracy (0.89, 0.85 and 0.84), while iCL2 was more difficult to classify (0.63). This approach will next be applied across the entire Brazilian soybean corridor, leading to an operational tool for monitoring the adoption of sustainable intensification practices recognized by Brazil’s Agriculture Low Carbon Plan (ABC PLAN).

Node classification is the task of inferring or predicting missing node attributes from information available for other nodes in a network. This paper presents a general prediction model to hierarchical multi-label classification, where the attributes to be inferred can be specified as a strict poset. It is based on a top-down classification approach that addresses hierarchical multi-label classification with supervised learning by building a local classifier per class. The proposed model is showcased with a case study on the prediction of gene functions for Oryza sativa Japonica, a variety of rice. It is compared to the Hierarchical Binomial-Neighborhood, a probabilistic model, by evaluating both approaches in terms of prediction performance and computational cost. The results in this work support the working hypothesis that the proposed model can achieve good levels of prediction efficiency, while scaling up in relation to the state of the art.

Monitoring of land cover plays an important role in effective environmental management, assessment of natural resources, environmental protection, urban planning and sustainable development. Increasing demand for accurate and repeatable information on land cover and land cover changes causes rapid development of the advanced, machine learning algorithms dedicated to land cover mapping using satellite images. Free and open access to Sentinel-2 data, characterized with high spatial and temporal resolution, increased the potential to map and to monitor land surface with high accuracy and frequency. Despite a considerable number of approaches towards land cover classification based on satellite data, there is still a challenge to clearly separate complex land cover classes, for example grasslands, arable land and wetlands. The aim of this study is to examine, whether a hierarchal classification of Sentinel-2 data can improve the accuracy of land cover mapping and delineation of complex land cover classes. The study is conducted in the Lodz Province, in central Poland. The pixel-based land cover classification is carried out using the machine learning Random Forest (RF) algorithm, based on a time series of Sentinel-2 imagery acquired in 2020. The following nine land cover classes are mapped: sealed surfaces, woodland broadleaved, woodland coniferous, shrubs, permanent herbaceous (grassy cover), periodically herbaceous (i.e., arable land), mosses, non-vegetated (bare soil) and water bodies. The land cover classification is conducted following two approaches: (1) flat, where all land cover classes are classified together, and (2) hierarchical, where the stratification is applied to first separate the most stable land cover classes and then classifying the most problematic once. The national databases served as the source of the reference sampling plots for the classification process. The process of selection and verification of the reference sampling plots is performed automatically. To assess the stability of the classification models the classification processes are performed iteratively. The results of this study confirmed that the hierarchical approach gave more accurate results compared to the commonly used flat approach. The median of the overall accuracy (OA) of the hierarchical classification was higher by 3–9 percentage points compared to the flat one. Of interest, the OA of the hierarchical classification reached 0.93–0.99, whereas the flat approach reached 0.90. Individual classes are also better classified in the hierarchical approach.

Herbarium sheets present a unique view of the world's botanical history, evolution, and biodiversity. This makes them an all–important data source for botanical research. With the increased digitization of herbaria worldwide and advances in the domain of fine–grained visual classification which can facilitate automatic identification of herbarium specimen images, there are many opportunities for supporting and expanding research in this field. However, existing datasets are either too small, or not diverse enough, in terms of represented taxa, geographic distribution, and imaging protocols. Furthermore, aggregating datasets is difficult as taxa are recognized under a multitude of names and must be aligned to a common reference. We introduce the Herbarium 2021 Half–Earth dataset: the largest and most diverse dataset of herbarium specimen images, to date, for automatic taxon recognition. We also present the results of the Herbarium 2021 Half–Earth challenge, a competition that was part of the Eighth Workshop on Fine-Grained Visual Categorization (FGVC8) and hosted by Kaggle to encourage the development of models to automatically identify taxa from herbarium sheet images.

Herbarium sheets present a unique view of the world's botanical history, evolution, and biodiversity. This makes them an all–important data source for botanical research. With the increased digitization of herbaria worldwide and advances in the domain of fine–grained visual classification which can facilitate automatic identification of herbarium specimen images, there are many opportunities for supporting and expanding research in this field. However, existing datasets are either too small, or not diverse enough, in terms of represented taxa, geographic distribution, and imaging protocols. Furthermore, aggregating datasets is difficult as taxa are recognized under a multitude of names and must be aligned to a common reference. We introduce the Herbarium 2021 Half–Earth dataset: the largest and most diverse dataset of herbarium specimen images, to date, for automatic taxon recognition. We also present the results of the Herbarium 2021 Half–Earth challenge, a competition that was part of the Eighth Workshop on Fine-Grained Visual Categorization (FGVC8) and hosted by Kaggle to encourage the development of models to automatically identify taxa from herbarium sheet images., Frontiers in Plant Science, 12, ISSN:1664-462X

Hierarchical classification has been previously demonstrated to yield more ac- curate results in comparison to flat classifiers; however, the tree-like structure intro- duces the problem of error propagation. Error propagation happens when a parent classifier within the hierarchy misclassifies an object. In traditional hierarchical structures, the system cannot recover from such misclassifications. In this paper, we propose a method of backtracking that allows the system to recover from such errors. For each non-root classifier in the structure, we add a new label option, “incorrect”. In the case that an object is classified as “incorrect” we perform back- tracking up the hierarchy and reclassify the object to the next best class from that previous level of the hierarchy. Through experimentation we have found that this method can help the system recover from individual instances of error propagation; however, more work is necessary to find an appropriate weight to give the “incorrect” class that doesn’t cause other errors, decreasing the over-all accuracy.

Due to different combinations of agriculture, livestock and forestry managed by rotation, succession and intercropping practices, integrated agriculture production systems such as integrated crop-livestock systems (iCL) constitute a very complex target and a challenge for automatic mapping of cropping practices based on remote sensing data. The overall objective of this study was to develop a classification strategy for the annual mapping of integrated Crop-Livestock systems (iCL) at a regional scale. This strategy was designed and tested in the six agro-climatic regions of Mato Grosso, the largest Brazilian soybean producer state, using MODIS satellite time-series images acquired between 2012 and 2019, ground data with heterogeneous distribution in space and time and a Random Forest classifier. The results showed that: 1. the use of unbalanced training samples with a class composition close to the real one was the right classifier training strategy; 2. the use of a single training database (pooling samples from different years and regions) to classify each region and year individually proved to be robust enough to provide similar classification accuracies in comparison to those based on the use of a database acquired for each region and for each year. The final hierarchical classification overall accuracy was 0.89 for Level 1, the cropping pattern level (single and double crops DC); 0.84 for Level 2, the DC category level (integrated system iCL soy-pasture/brachiaria, soy-cotton and soy-cereal); 0.77 for Level 3, the iCL level (iCL1 soy-pasture and iCL2 soy-pasture mixed with corn). The F-scores for DC, iCL and iCL1 cropping systems presented high accuracy (0.89, 0.85 and 0.84), while iCL2 was more difficult to classify (0.63). This approach will next be applied across the entire Brazilian soybean corridor, leading to an operational tool for monitoring the adoption of sustainable intensification practices recognized by Brazil's Agriculture Low Carbon Plan (ABC PLAN). Made available in DSpace on 2022-04-04T15:00:45Z (GMT). No. of bitstreams: 1 Monitoring-complex-integrated-crop-livestock-systems-at-regional-scale-in-Brazil-2022.pdf: 8358810 bytes, checksum: 91d7d9fad47d513abb47458e709f1d4c (MD5) Previous issue date: 2022

When merging existing similar datasets, it would be attractive to benefit from a higher detection rate of objects and the additional partial ground-truth samples for improving object classification. To this end, a novel CNN detector with a hierarchical binary classification system is proposed. The detector is based on the Single-Shot multibox Detector (SSD) and inspired by the hierarchical classification used in the YOLO9000 detector. Localization and classification are separated during training, by introducing a novel loss term that handles hierarchical classification in the loss function (SSD-ML). We experiment with the proposed SSD-ML detector on the generic PASCAL VOC dataset and show that additional super-categories can be learned with minimal impact on the overall accuracy. Furthermore, we find that not all objects are required to have classification label information as classification performance only drops from 73.3 % to 70.6 % while 60 % of the label information is removed. The flexibility of the detector with respect to the different levels of details in label definitions is investigated for a traffic surveillance application, involving public and proprietary datasets with non-overlapping class definitions. Including classification label information from our dataset raises the performance significantly from 70.7 % to 82.2 %. The experiments show that the desired hierarchical labels can be learned from the public datasets, while only using box information from our dataset. In general, this shows that it is possible to combine existing datasets with similar object classes and partial annotations and benefit in terms of growth of detection rate and improved class categorization performance.

Public procurement accounts for a 14% of the annual budget of the different governments of the European Union. In Europe, contracting processes are classified using Common Procurement Vocabulary codes (CPVs), a taxonomy designed to facilitate statistical reporting, search and the creation of alerts that can be used by potential bidders. CPVs are commonly assigned manually by public employees in charge of contracting processes. However, CPV classification is not a trivial task, as there are more than 9,000 different CPV categories, which are often assigned following heterogeneous criteria. In this paper we have created a CPV classifier that uses as an input the textual description of the contracting process, and assigns CPVs from the 45 top-level CPV categories. We work only with texts in Spanish, although our approach may be easily extended to other languages. Our results improve the state of the art (10% F1-score improvement) and are available online. Las licitaciones públicas suponen el 14% del presupuesto anual de la Unión Europea. En Europa, los procesos de contratación se clasifican usando la taxonomía Common Procurement Vocabulary (CPVs), diseñada para facilitar la generación de estadísticas, las búsquedas y la creación de alertas que puedan utilizar los posibles licitadores. Los códigos CPV suelen ser asignados manualmente por los empleados públicos encargados del proceso de contratación. Sin embargo, la clasificación de textos de acuerdo con estos códigos no es trivial, pues existen más de 9000 CPVs y no siempre se siguen los mismos criterios para su asignación. En este artículo se propone un clasificador que utiliza como entrada la descripción textual del proceso de contratación, y produce códigos de entre las 45 categorías de CPV más generales de la jerarquía. Trabajamos solo con textos en español, aunque nuestro enfoque puede extenderse fácilmente a otros idiomas. Los resultados obtenidos superan el estado del arte (10% de mejora en F1), y se encuentran disponibles online. This work has been supported by NextProcurement European Action (grant agreement INEA/CEF/ICT/A2020/2373713-Action 2020-ES-IA-0255) and the Madrid Government (Comunidad de Madrid-Spain) under the Multiannual Agreement with Universidad Politécnica de Madrid in the line Support for R&D projects for Beatriz Galindo researchers, in the context of the V PRICIT (Regional Programme of Research and Technological Innovation).

Altres ajuts: CERCA Programme/Generalitat de Catalunya ; ACCIO agency to CVC's general activities Recent works have made significant progress in novelty detection, i.e., the problem of detecting samples of novel classes, never seen during training, while classifying those that belong to known classes. However, the only information this task provides about novel samples is that they are unknown. In this work, we leverage hierarchical taxonomies of classes to provide informative outputs for samples of novel classes. We predict their closest class in the taxonomy, i.e., its parent class. We address this problem, known as hierarchical novelty detection, by proposing a novel loss, namely Hierarchical Cosine Loss that is designed to learn class prototypes along with an embedding of discriminative features consistent with the taxonomy. We apply it to traffic sign recognition, where we predict the parent class semantics for new types of traffic signs. Our model beats state-of-the art approaches on two large scale traffic sign benchmarks, Mapillary Traffic Sign Dataset (MTSD) and Tsinghua-Tencent 100K (TT100K), and performs similarly on natural images benchmarks (AWA2, CUB). For TT100K and MTSD, our approach is able to detect novel samples at the correct nodes of the hierarchy with 81% and 36% of accuracy, respectively, at 80% known class accuracy.

In this work, freshwater microplastic samples collected from four different stations along the Italian Po river were characterized in terms of abundance, distribution, category, morphological and morphometrical features, and polymer type. The correlation between microplastic category and polymer type was also evaluated. Polymer identification was carried out developing and implementing a new and effective hierarchical classification logic applied to hyperspectral images acquired in the short-wave infrared range (SWIR: 1000–2500 nm). Results showed that concentration of microplastics ranged from 1.89 to 8.22 particles/m3, the most abundant category was fragment, followed by foam, granule, pellet, and filament and the most diffused polymers were expanded polystyrene followed by polyethylene, polypropylene, polystyrene, polyamide, polyethylene terephthalate and polyvinyl chloride, with some differences in polymer distribution among stations. The application of hyperspectral imaging (HSI) as a rapid and non-destructive method to classify freshwater microplastics for environmental monitoring represents a completely innovative approach in this field.

Hierarchical multi-label classification is a challenging task implying the encoding of a high level constraint in the neural network model. Before the rise of this field, the classification was done without paying attention to the hierarchical links existing between data. Nevertheless, information relating the classes and subclasses may be very useful for improving the network performances. Recently, some works have integrated the hierarchy information by proposing new neural network architectures (called B-CNN or H-CNN), achieving promising results. However with these architectures, the network is separated into blocks where each block is responsible for predicting only the classes of a given level in the hierarchy. In this paper, we propose a novel architecture such that the whole network layers are involved in the prediction of the entire labels of a sample, i.e., from its class in the top level of the hierarchy to its class in the bottom level. The proposed solution is based on a Bayesian adjustment encoding the hierarchy in terms of conditional probabilities, together with a customized semantic loss function that penalizes drastically the hierarchy violation. A teacher forcing strategy learning is used to enhance the learning quality. Thanks to this approach, we could outperform the state of the art results in terms of accuracy (improved for all levels) and also in terms of hierarchy coherence.

International audience; We present in this paper a new classification approach for identifying questions during human-machine interactions and more specifically in dialogue systems. The difficulty in this task is first to be domainindependent, reusable whatever the dialogue application and second to be capable of a real time processing, in order to fit with the needs of reactivity in dialogue systems. The task is then different than that of question classification usually addressed in question-answering systems. We propose in this paper a hierarchical classifier in two steps, filtering first question/no-question utterances and second the type of the question. Our method reaches a f-score of 98% for the first step and 97% for the second one, representing the state of the art for this task.

Image registration is the process of aligning images by finding the spatial relation between the images. Assuming two images called fixed and moving images are taken at different time, different spatial location, or via a different imaging technique, the aim of image registration is to find an optimal transformation that aligns the fixed and the moving images. Performing an automatic fast image registration with less manual finetuning can speed up numerous medical image processing procedures. In addition, an automatic quality assessment of registration can speed up this time-consuming task. In this thesis, we developed a fast learning-based image registration technique called RegNet.Predicting registration error can be useful for evaluation of registration procedures, which is important for the adoption of registration techniques in the clinic. In addition, quantitative error prediction can be helpful in improving the registration quality. In this thesis, we proposed two quality assessment mechanisms using random forests (RF) and convolutional long short term memory (ConvLSTM), in which the latter performs faster and more accurate.

Accurate and efficient individual tree species (ITS) classification is the basis of fine forest resource management. It is a challenge to classify individual tree species in dense forests using remote sensing imagery. In order to solve this problem, a new ITS classification method was proposed in this study, in which a hierarchical convolutional neural network (H-CNN) model and multi-temporal high-resolution Google Earth images were employed. In an experiment conducted in a forest park in Beijing, China, GE images of several significant phenological phases of broad-leaved forests, namely, before and after the mushrooming period, the growth period, and the wilting period, were selected, and ITS classifications based on these images along with several typical CNN models and the H-CNN model were conducted. In the experiment, the classification accuracy of the multitemporal images was higher by 7.08–12.09% than those of the single-temporal images, and the H-CNN model offered an OA accuracy 2.66–3.72% higher than individual CNN models, demonstrating that multitemporal images rich in the phenological features of individual tree species, together with a hierarchical CNN model, can effectively improve ITS classification.

International audience; When soil remediation specialists clean up a new site, they have a long time manually revising digital reports previously written by other experts, where they look for necessary information in accordance with similar characteristics of polluted fields. Important information lies in tables, graphs, maps, drawings and their associated captions. Therefore, experts have to be able to quickly access these content-rich elements, instead of manually scrolling through each page of entire reports. Since this information is multimodal (image and text) and follows a semantically hierarchical structure, we propose a classification algorithm that takes these two constraints into account. In contrast to existing works using either multimodal system or hierarchical classification model, we explore the combination of state-of-the-art methods from multimodal systems (image and text modalities) and hierarchical classification systems. By this combination, we tackle the constraints of our classification process: small dataset, missing modalities, noisy data, and non-English corpus. Our evaluation shows that the multimodal hierarchical system outperforms the unimodal and that the performance of multimodal system with a joint combination of hierarchical classification and flat classification on different modalities provides promising results.

In this paper, we propose a novel method for plant leaves recognition by incorporating an unsupervised convolutional auto-encoder (CAE) and Siamese neural network in a unified framework by considering Siamese as an alternative to the conventional loss of CAE. Rather than the conventional exploitation of CAE and Siamese, in our case we have proposed to extend CAE for a novel supervised scenario by considering it as one-class learning classifier. For each class, CAE is trained to reconstruct its positive and negative examples and Siamese is trained to distinguish the similarity and the dissimilarity of the obtained examples. On the contrary and asymmetric to the related hierarchical classification schemes which require pre-knowledge on the dataset being recognized, we propose a hierarchical classification scheme that doesn’t require such a pre-knowledge and can be employed by non-experts automatically. We cluster the dataset to assemble similar classes together. A test image is first assigned to the nearest cluster, then matched to one class from the classes that fall under the determined cluster using our novel one-class learning classifier. The proposed method has been evaluated on the ImageCLEF2012 dataset. Experimental results have proved the superiority of our method compared to several state-of-the art methods.

Image classification has been a pivotal area of research in Deep Learning, with a vast body of literature working to tackle the problem, constantly striving to achieve higher accuracies. This push to reach achieve greater prediction accuracy however, has further exacerbated the black box phenomenon which is inherent of neural networks, and more for so CNN style deep architectures. Likewise, it has lead to the development of highly tuned methods, suitable only for a specific data sets, requiring significant work to alter given new data. Although these models are capable of producing highly accurate predictions, we have little to no ability to understand the decision process taken by a network to reach a conclusion. This factor poses a difficulty in use cases such as medical diagnostics tools or autonomous vehicles, which require insight into prediction reasoning to validate a conclusion or to debug a system. In essence, modern applications which utilize deep networks are able to learn to produce predictions, but lack interpretability and a deeper understanding of the data. Given this key point, we look to decision trees, opposite in nature to deep networks, with a high level of interpretability but a low capacity for learning. In our work we strive to merge these two techniques as a means to maintain the capacity for learning while providing insight into the decision process. More importantly, we look to expand the understanding of class relationships through a tree architecture. Our ultimate goal in this work is to create a technique able to automatically create a visual feature based knowledge hierarchy for class relations, applicable broadly to any data set or combination thereof. We maintain these goals in an effort to move away from specific systems and instead toward artificial general intelligence (AGI). AGI requires a deeper understanding over a broad range of information, and more so the ability to learn new information over time. In our work we embed networks of varying sizes and complexity within decision trees on a node level, where each node network is responsible for selecting the next branch path in the tree. Each leaf node represents a single class and all parent and ancestor nodes represent groups of classes. We designed the method such that classes are reasonably grouped by their visual features, where parent and ancestor nodes represent hidden super classes. Our work aims to introduce this method as a small step towards AGI, where class relations are understood through an automatically generated decision tree (representing a class hierarchy), capable of accurate image classification. Master of Science Many modern day applications make use of deep networks for image classification. Often these networks are incredibly complex in architecture, and applicable only for specific tasks and data. Standard approaches use just a neural network to produce predictions. However, the internal decision process of the network remains a black box due to the nature of the technique. As more complex human related applications, such as medical image diagnostic tools or autonomous driving software, are being created, they require an understanding of reasoning behind a prediction. To provide this insight into the prediction reasoning, we propose a technique which merges decision trees and deep networks. Tested on the MNIST image data set we were able to achieve an accuracy over 99.0%. We were also able to achieve an accuracy over 73.0% on the CIFAR-10 image data set. Our method is found to create decision trees that are easily understood and are reasonably capable of image classification.

We introduce and study knowledge drift (KD), a complex form of drift that occurs in hierarchical classification. Under KD the vocabulary of concepts, their individual distributions, and the is-a relations between them can all change over time. The main challenge is that, since the ground-truth concept hierarchy is unobserved, it is hard to tell apart different forms of KD. For instance, introducing a new is-a relation between two concepts might be confused with individual changes to those concepts, but it is far from equivalent. Failure to identify the right kind of KD compromises the concept hierarchy used by the classifier, leading to systematic prediction errors. Our key observation is that in many human-in-the-loop applications (like smart personal assistants) the user knows whether and what kind of drift occurred recently. Motivated by this, we introduce TRCKD, a novel approach that combines automated drift detection and adaptation with an interactive stage in which the user is asked to disambiguate between different kinds of KD. In addition, TRCKD implements a simple but effective knowledge-aware adaptation strategy. Our simulations show that often a handful of queries to the user are enough to substantially improve prediction performance on both synthetic and realistic data., 8 pages, figures 4, includes supplementary material (3 pages, 4 figures), code: https://gitlab.com/abonte/handling-knowledge-drift

Cotton seed production is the main form of agriculture in Xinjiang, China. Unreasonable distribution of cotton seed production results in a waste of water, land, and human resources. In this study, we established a hierarchical classification integrating method, investigated the spatial suitability of climate, land and water resources, and infrastructure, examined the production risk and planting history, and integrated spatial suitability and production risk and history to produce the spatial layout of seed production for early-maturing cotton (EMC), early–medium-maturing cotton (EMMC), and long staple cotton (LSC) in Xinjiang. The results indicated that the appropriate areas for EMC, EMMC, and LSC seed production are 6.4 × 105, 5.5 × 105, and 3.6 × 105 ha, respectively. By combining the suitable areas of seed production for the three cotton species, we concluded that the superior and most suitable area for producing cotton seed of EMC, EMMC, or LSC is located in the western Tarim Basin. The sub-suitable area for cotton seed production of EMC, EMMC, or LSC is mostly distributed in the western and northern Tarim Basin. This research provides a good solution to the selection of cotton seed production base, however, adoption will depend on the actual preference and market factors.

Graphical abstract, Today machine learning methods are commonly deployed for bacterial species identification using MALDI-TOF mass spectrometry data. However, most of the studies reported in literature only consider very traditional machine learning methods on small datasets that contain a limited number of species. In this paper we present benchmarking results on an unprecedented scale for a wide range of machine learning methods, using datasets that contain almost 100,000 spectra and more than 1000 different species. The size and the diversity of the data allow to compare three important identification scenarios that are often not distinguished in literature, i.e., identification for novel biological replicates, novel strains and novel species that are not present in the training data. The results demonstrate that in all three scenarios acceptable identification rates are obtained, but the numbers are typically lower than those reported in studies with a more limited analysis. Using hierarchical classification methods, we also demonstrate that taxonomic information is in general not well preserved in MALDI-TOF mass spectrometry data. For the novel species scenario, we apply for the first time neural networks with Monte Carlo dropout, which have shown to be successful in other domains, such as computer vision, for the detection of novel species.

La importancia de representar los ecosistemas neotropicales surge de la necesidad de generar cartografía que facilite comprender su segregación en el espacio. El objetivo de nuestro trabajo fue diseñar un modelo de representación cartográfica de ecosistemas, a través de procesos que integran los insumos cartográficos y sus respectivas escalas, asociados a un sistema de clasificación jerárquica. Para ello, se determinaron niveles jerárquicos con base en criterios geofísicos y biofísicos, característicos de los ecosistemas neotropicales, se seleccionaron insumos cartográficos en cada nivel jerárquico asociado a su respectiva escala. Se concluye que los ecosistemas neotropicales deben ser representados a mayores escalas, ya que es la única forma de obtener el detalle necesario de los atributos que los caracterizan, esta necesidad está determinada por la complejidad derivada de la diversidad geofísica y biológica de esta región. Provides cartography to represent neotropicals ecosystems helps to understand segregation in space. When mapping ecosystems, ecologists have some problems, some are related to scales and cartographic inputs. The goal of our work was to design a model of cartographic representation of ecosystems, through processes that integrate cartographic inputs and their respective scales, associated with a hierarchical classification system. For this purpose, we determine hierarchical levels based on geophysical and biophysical criteria, characteristic of Neotropical ecosystems. We select cartographic inputs in each hierarchical level associated with their respective scale. We conclude that the Neotropical ecosystems must be represented at greater scales, since it is the only way to obtain the necessary detail of the attributes that characterize them, this need is determined by the complexity derived from the geophysical and biological diversity of this region. 198-214 dyanezc@uteq.edu.ec xandradeprot@gmail.com renoharobull@gmail.com Semestral http://www.saber.ula.ve/regeoven/

Food safety policies have gained considerable importance in recent years, food safety being one of the indicators that illustrates the standard of living and quality of life within a nation state. In order to assess food security, we analyzed the Global Food Security Index (GFSI) that, economically, is based on GDP, degree of poverty and agricultural production, extending also to areas such as government and public policies, which are usually not directly included in food safety indicators or generating factors. Considering the importance and impact of food safety, starting from the theoretical concepts, from the current state of food safety in Europe, presented by the means of GFSI and its components, we developed a set of measures based on the grouping (correlation) of states following the analysis of hierarchical ranking of clusters. To analyse clusters hierarchical ranking the food safety indicators were used as input, dependent data. As an independent variable, with a strong influence on all the others, determined by means of multiple linear regression, we considered GDP/inhabitant at the level of each analysed state. The design of the set of measures considered the correlation that can be established among the various GFSI indicators that influence and generate the current state of food security in different European countries and the influence these indicators can have on maintaining or improving this state.

Bu çalışmanın amacı dörtgenlerin içeren ve hariç tutan tanımlarına odaklanılarak yapılan öğretimin, 7. sınıf öğrencilerinin dörtgenleri ilişkilendirme becerilerine etkisini incelemektir. Araştırmanın örneklemini Orta Karadeniz Bölgesinde il merkezine bağlı bir köy okulunda öğrenim gören; deney grubunda 16, kontrol grubunda 15 öğrenci olmak üzere toplam 31 tane 7. sınıf öğrencisi oluşturmaktadır. Çalışmada zayıf deneysel desenlerden ön-test son-test kontrol gruplu desen kullanılmıştır. Veri toplama aracı olarak araştırmacılar tarafından alan yazında kullanılan sorular derlenerek, uzman görüşü rehberliğinde hazırlanan, dörtgenler erişi testi kullanılmıştır. Öğretim etkinlikleri 5 hafta boyunca devam etmiştir. Dörtgenler konusu, deney grubundaki öğrencilere içeren tanımlara, kontrol grubunda ise hariç tutan tanımlara odaklanılarak anlatılmıştır. Bulgulara göre, ön test puanları arasında deney grubu ve kontrol grubu arasında başarı açısından istatistiksel olarak anlamlı bir fark yokken, son test puanları arasında deney grubu lehine istatistiksel olarak anlamlı bir fark olduğu saptanmıştır. Sonuç olarak dörtgenlerin içeren tanımlarına odaklanılarak yapılan öğretim etkinliklerinin öğrenci başarısı üzerinde olumlu etkileri vardır denilebilir.

To solve the problem of low efficiency, the complexity of the interactive operation, and the high degree of manual intervention in existing methods, we propose a novel approach based on the sparse voxel octree and 3D convolution neural networks (CNNs) for segmenting and classifying tooth types on the 3D dental models. First, the tooth classification method capitalized on the two-level hierarchical feature learning is proposed to solve the misclassification problem in highly similar tooth categories. Second, we exploit an improved three-level hierarchical segmentation method based on the deep convolution features to conduct segmentation of teeth-gingiva and inter-teeth, respectively, and the conditional random field model is used to refine the boundary of the gingival margin and the inter-teeth fusion region. The experimental results show that the classification accuracy in Level_1 network is 95.96%, the average classification accuracy in Level_2 network is 88.06%, and the accuracy of tooth segmentation is 89.81%. Compared with the existing state-of-the-art methods, the proposed method has higher accuracy and universality, and it has great application potential in the computer-assisted orthodontic treatment diagnosis.

In this paper, we address fish species identification in underwater video for marine monitoring applications such as the study of marine biodiversity. Video is the least disruptive monitoring method for fish but requires efficient techniques of image processing and analysis to overcome challenging underwater environments. We propose two Deep Convolutional Neural Network (CNN) approaches for fish species classification in unconstrained underwater environment. In the first approach, we use a traditional transfer learning framework and we investigate a new technique based on training/validation loss curves for targeted data augmentation. In the second approach, we propose a hierarchical CNN classification to classify fish first into family levels and then into species categories. To demonstrate the effectiveness of the proposed approaches, experiments are carried out on two benchmark datasets for automatic fish identification in unconstrained underwater environment. The proposed approaches yield accuracies of 99.86% and 81.53% on the Fish Recognition Ground-Truth dataset and LifeClef 2015 Fish dataset, respectively.

Human sensing is concerned with techniques for inferring information about humans from various sensing modalities. Examples of human sensing applications include human activity (or action) recognition, emotion recognition, tracking and localisation, identification, presence and motion detection, occupancy estimation, gesture recognition, and breath rate estimation. The first question addressed in this thesis is whether micro or macro models are a better design choice for human sensing systems. Micro models are models exclusively trained with data from a single entity, such as a Wi-Fi link, user, or other identifiable data-generating component. We consider micro and macro models in two human sensing applications, viz. Human Activity Recognition (HAR) from wearable inertial sensor data and device-free human presence detection from Wi-Fi signal data. The HAR literature is dominated by person-independent macro models. The few empirical studies that consider both micro and macro models evaluate them with either only one data-set or only one HAR algorithm, and report contradictory results. The device-free sensing literature is dominated by link-specific micro models, and the few papers that do use macro models do not evaluate their micro counterparts. Given the little and contradictory evidence, it remains an open question whether micro or macro models are a better design choice. We evaluate person-specific micro and person-independent macro models across seven HAR benchmark data-sets and four learning algorithms. We show that person-specific models (PSMs) significantly outperform the corresponding person-independent model (PIM) when evaluated with known users. To apply PSMs to data from new users, we propose ensembles of PSMs, which are improved by weighting their constituent PSMs according to their performance on other training users. We propose link-specific micro models to detect human presence from ambient Wi-Fi signal data. We select a link-specific model from the available training links, and show that this approach outperforms multi-link macro models. The second question addressed in this thesis is whether human sensing methods can be improved with domain knowledge. Specifically, we propose expert hierarchies (EHs) as an intuitive way to encode domain knowledge and simplify multi-class HAR, without negatively affecting predictive performance. The advantages of EHs are that they have lower time complexity than domain-agnostic methods and that their constituent classifiers are statistically independent. This property enables targeted tuning, and modular and iterative development of increasingly fine-grained HAR. Although this has inspired several uses of domain-specific hierarchical classification for HAR applications, these have been ad-hoc and without comparison to standard domain-agnostic methods. Therefore, it remains unclear whether they carry a penalty on predictive performance. We design five EHs and compare them to the best-known domain-agnostic methods. Our results show that EHs indeed can compete with more popular multi-class classification methods, both on the original multi-class problem and on the EHs' topmost levels.

There are several local indicators of spatial association (LISA) that allow exploration of local patterns in spatial data. Despite numerous situations where categorical variables are encountered, few attempts have been devoted to the development of methods to explore the local spatial pattern in categorical data. To our knowledge, there is no indicator of local spatial association that can be used for both continuous and categorical data. We introduce ELSA, which can be used for exploring and testing local spatial association for continuous and categorical variables. We provide the R-package elsa for making these computations.

X Congreso Geológico de España, 5-7 Julio 2021, Vitoria - Gasteiz, La estación sísmica CANF, integrada en el proyecto Geodyn del Laboratorio Subterráneo Canfranc (LSC), se localiza en la zona central de los Pirineos, en un emplazamiento subterráneo cercano al lecho del río Aragón, por lo que ofrece la oportunidad de verificar la viabilidad del uso de datos sísmicos para monitorizar eventos hidrológicos. En esta contribución nos centramos en la identificación de señales sísmicas generadas por variaciones en la descarga del río debidas al deshielo, a lo largo de un período de seis años (2011-2016). Durante los episodios de deshielo, las variaciones en la descarga en el río de drenaje generan señales sísmicas con características específicas que permiten su discriminación de otras fuentes de vibraciones. Hemos desarrollado una metodología basada en la clasificación jerárquica de los espectros sísmicos para identificar los días en los que se produce el deshielo y agruparlos en función de la intensidad de la descarga, identificando hasta cuatro tipos diferentes de episodios. Estos resultados demuestran que los datos sísmicos pueden proporcionar una buena herramienta para monitorizar los procesos de deshielo a largo plazo, contribuyendo así a los estudios sobre el cambio climático., The CANF seismic broad-band station, part of the Geodyn facility in the Laboratorio Subterráneo de Canfranc (LSC), is located in an underground site in the Central Pyrenees, close to the Aragón River channel, and hence offers a good opportunity to study the long-term performance of the seismic monitoring of hydrological events. We focus here on the identification of seismic signals generated by variations in river discharge due to snow melting over six years (2011-2016). During snowmelt episodes, the temporal variations of the discharge in the drainage river result in characteristic seismic signals that can be discriminated from other sources of background vibrations. We have developed a methodology based on the analysis of the seismic spectra in order to identify days with active thaw, hence allowing to monitor the time occurrence and intensity of the snowmelting stages. A hierarchical classification of the seismic spectra has allowed us to group active thaw days according to the intensity of the discharge generated by the melting process, identifying up to four different types of episodes. These results prove that seismic data can provide a good tool to monitor snowmelt on a long-term scale, hence contributing to climate change studies., This is a contribution of the MISTERIOS project [CGL2013-48601-C2-1-R] funded by the Spanish Ministry of Economy, Industry and Competitiveness. Additional support from the Generalitat de Catalunya grant 2017SGR1022 and from the SANIMS project (RTI2018-095594-B-I00), funded by the Spanish Ministry of Science, Innovation and Universities.

This paper formulates the problem of predictive maintenance for complex systems as a hierarchical multi-class classification task. This formulation is useful for equipment with multiple sub-systems and components performing heterogeneous tasks. Often, the data available describes the whole system's operation and is not ideal for accurate condition monitoring. In this setup, specialized predictive models analyzing one component at a time rarely perform much better than random. However, using machine learning and hierarchical approaches, we can still exploit the data to build a fault isolation system that provides measurable benefits for technicians in the field. We propose a method for creating a taxonomy of components to train hierarchical classifiers that aim to identify the faulty component. The output of this model is a structured set of predictions with different probabilities for each component. In this setup, traditional machine learning metrics fail to capture the relationship between the performance of the models and its usefulness in the field.We introduce a new metric to evaluate our approach's benefits; it measures the number of tests a technician needs to perform before pinpointing the faulty component. Using a dataset from a real-case problem coming fro the automotive industry, we demonstrate how traditional machine learning performance metrics, like accuracy, fail to capture practical benefits. Our proposed hierarchical approach succeeds in exploiting the information in the data and outperforms non-hierarchical machine learning solutions. In addition, we can identify the weakest link of our fault isolation model, allowing us to improve it efficiently.

International audience; Aim Dynamic global vegetation models (DGVMs) use a discretization of forest vegetation based on plant functional types (PFTs). The physiological and ecological parameters used to model a given PFT are usually fixed, being defined from point-based observations, while model applications are often grid-based. This rigid approach causes spatial biases in the results of DGVMsimulated productivity and biomass-related variables. We aim to overcome this limitation with a new approach that uses a hierarchical classification of forest PFT parameters from traits retrieved from the literature and from the TRY global database of plant traits. This approach is applied to temperate conifers in the Orchidee-FM DGVM, which has previously been shown to produce systematic biases in the simulation of biomass and biomass increments. Location Temperate coniferous forests in France. Methods The five major coniferous species in France, Abies alba, Picea abies, Pinus pinaster, Pinus sylvestris and Pseudotsuga menziesii, were grouped objectively into PFTs within the Orchidee-FM DGVM using a hierarchical classification based on 12 key attributes related to photosynthesis, phenology and allometric relationships. Results We show that the single PFT covering all temperate coniferous forests used by default in Orchidee-FM could be replaced by two representative subcategories defined by grouping species-level data without necessarily having to adopt a set of parameters for each species. The definition of new temperate conifer PFTs with this approach allows us to reduce the spatial heterogeneity by 40% on average in model–Measurement misfit for stand volume, growth and stand density at the regional scale. Main conclusions The proposed approach to improve the representation of PFTs in DGVMs, while keeping the number of different PFTs manageable, is promising for application to regions where a single PFT can correspond to a number of different species.

In this paper we propose a supervised method to predict registration misalignment using convolutional neural networks (CNNs). This task is casted to a classification problem with multiple classes of misalignment: “correct” 0-3 mm, “poor” 3-6 mm and “wrong” over 6 mm. Rather than a direct prediction, we propose a hierarchical approach, where the prediction is gradually refined from coarse to fine. Our solution is based on a convolutional Long Short-Term Memory (LSTM), using hierarchical misalignment predictions on three resolutions of the image pair, leveraging the intrinsic strengths of an LSTM for this problem. The convolutional LSTM is trained on a set of artificially generated image pairs obtained from artificial displacement vector fields (DVFs). Results on chest CT scans show that incorporating multi-resolution information, and the hierarchical use via an LSTM for this, leads to overall better F1 scores, with fewer misclassifications in a well-tuned registration setup. The final system yields an accuracy of 87.1%, and an average F1 score of 66.4% aggregated in two independent chest CT scan studies.

In this paper we propose a supervised method to predict registration misalignment using convolutional neural networks (CNNs). This task is casted to a classification problem with multiple classes of misalignment: "correct" 0-3 mm, "poor" 3-6 mm and "wrong" over 6 mm. Rather than a direct prediction, we propose a hierarchical approach, where the prediction is gradually refined from coarse to fine. Our solution is based on a convolutional Long Short-Term Memory (LSTM), using hierarchical misalignment predictions on three resolutions of the image pair, leveraging the intrinsic strengths of an LSTM for this problem. The convolutional LSTM is trained on a set of artificially generated image pairs obtained from artificial displacement vector fields (DVFs). Results on chest CT scans show that incorporating multi-resolution information, and the hierarchical use via an LSTM for this, leads to overall better F1 scores, with fewer misclassifications in a well-tuned registration setup. The final system yields an accuracy of 87.1%, and an average F1 score of 66.4% aggregated in two independent chest CT scan studies.

This thesis investigates how pre-trained contextualized language models can be adapted for multi-label text classification of Swedish news articles. Various classifiers are built on pre-trained BERT and ELECTRA models, exploring global and local classifier approaches. Furthermore, the effects of domain specialization, using additional metadata features and model compression are investigated. Several hundred thousand news articles are gathered to create unlabeled and labeled datasets for pre-training and fine-tuning, respectively. The findings show that a local classifier approach is superior to a global classifier approach and that BERT outperforms ELECTRA significantly. Notably, a baseline classifier built on SVMs yields competitive performance. The effect of further in-domain pre-training varies; ELECTRA’s performance improves while BERT’s is largely unaffected. It is found that utilizing metadata features in combination with text representations improves performance. Both BERT and ELECTRA exhibit robustness to quantization and pruning, allowing model sizes to be cut in half without any performance loss.