Selecting and Combining Classifiers Based on Centrality Measures

Centrality measures have been helping to explain the behavior of objects, given their relation, in a wide variety of problems, since sociology to chemistry. This work considers these measures to assess the importance of every classifier belonging to an ensemble of classifiers, aiming to improve a Multiple Classifier System (MCS). Assessing the classifier’s importance by employing centrality measures, inspired two different approaches: one for selecting classifiers and another for fusion. The selection approach, called Centrality Based Selection (CBS), adopts a trade-off between the classifier’s accuracy and their diversity. The sub-optimal selected subset presents good results against selection methods from the literature, being superior in 67.22% of the cases. The second approach, the integration, is named Centrality Based Fusion (CBF). This approach is a weighted combination method, which is superior to literature in 70% of the cases.

Augmenting Deep Learning Performance in an Evidential Multiple Classifier System

The main objective of this work is to study the applicability of ensemble methods in the context of deep learning with limited amounts of labeled data. We exploit an ensemble of neural networks derived using Monte Carlo dropout, along with an ensemble of SVM classifiers which owes its effectiveness to the hand-crafted features used as inputs and to an active learning procedure. In order to leverage each classifier’s respective strengths, we combine them in an evidential framework, which models specifically their imprecision and uncertainty. The application we consider in order to illustrate the interest of our Multiple Classifier System is pedestrian detection in high-density crowds, which is ideally suited for its difficulty, cost of labeling and intrinsic imprecision of annotation data. We show that the fusion resulting from the effective modeling of uncertainty allows for performance improvement, and at the same time, for a deeper interpretation of the result in terms of commitment of the decision.

Genders and classifiers in Kampa (Arawak) languages of Peru

This chapter examines the semantics, morphosyntax, and functions of the gender and classifier systems of Kampa Arawak languages of Peru. All Kampa languages have genders and classifiers. Their origin and diachronic development are different. Gender agreement morphology comes from the pan-Kampa verbal person markers. The sources of multiple classifiers are bound nouns inflected on the pattern of obligatorily possessed nouns, unbound nouns, and bound verb roots; these are considered in the context of compounding and noun incorporation. Gender marking is mandatory and exhaustive, being reflected in the agreement marking on noun modifiers (adjectives, demonstratives, possessor NP), possessive pronouns, demonstrative identifiers, personal pronouns, verbs, and coordinating operators. Multiple classifiers show less exuberant distribution, occurring on nouns, verbs, number words, and adjectives. Classifiers are neither sensitive to gender nor animacy. Classifiers are semantically motivated, showing semantic agreement with controller nouns. The multiple classifier system does not participate in syntactic parsing of constituents via morphological agreement. The main purpose of their use is pragmatic.

LAND-COVER MAPS USING MULTIPLE CLASSIFIER SYSTEM FOR POST-DISASTER LANDSCAPE MONITORING

<p><strong>Abstract.</strong> Recently, land cover maps created from high resolution satellite images have been used for landscape analysis, in order to understand the impact of natural disasters on biodiversity and ecosystems. Conventional land cover classification methods, however, suffer from problems with isolated pixels (salt and pepper effect). Filtering can remove the isolated pixels, but can also result in loss of accurate information. The purpose of this study is to create a land cover map for landscape analysis of large-scale disturbances caused by the Great East Japan Earthquake of 2011, utilizing a Multiple Classifier System (MCS), which allows for reduction of isolated pixels while maintaining classification accuracy. RapidEye satellite images covering the Pacific Ocean side of the Tohoku district damaged by the earthquake and subsequent tsunami were obtained for 2010, 2011, 2012 and 2016, and land cover classification was implemented using individual classifiers and the MCS method. The results showed that the MCS land cover map was able to reduce the number of isolated pixels significantly (61-71%) compared with the individual classifiers, while maintaining very high accuracy (0.976-0.986) for all four years. These results indicate that MCS land cover maps have a great potential for analyzing disturbances following infrequent largescale natural disasters such as earthquakes and tsunami, and for monitoring the process of recovery afterwards. We expect that the results of this research will be useful in managing the recovery process in the region disturbed by the Great Eastern Japan Earthquake and Tsunami of 2011, and also for developing future Ecosystem-based Disaster Risk Reduction programs for the region.</p>