As AI systems become more capable, it is important that their decisions are understandable and aligned with human expectations. A key challenge is the lack of interpretability in deep models. Existing methods such as GradCAM generate heatmaps but provide limited conceptual insight, while prototype-based approaches offer example-based explanations but often rely on rigid region selection and lack semantic consistency. To address these limitations, we propose PCMNet, a Part-Prototypical Concept Mining Network that learns human-comprehensible prototypes from meaningful regions without extra supervision. By clustering these into concept groups and extracting concept activation vectors, PCMNet provides structured, concept-level explanations and enhances robustness under occlusion and adversarial conditions, which are both critical for building reliable and aligned AI systems. Experiments across multiple benchmarks show that PCMNet outperforms state-of-the-art methods in interpretability, stability, and robustness. This work contributes to AI alignment by enhancing transparency, controllability, and trustworthiness in modern AI systems.

Offline Handwritten Signature Verification (offline HSV) analyzes static images of signatures to distinguish between genuine and forged samples. To create training data for such systems, negative samples, commonly known as random forgeries, are typically drawn from genuine signatures of other users, as real-world datasets often lack actual forgeries. While this strategy helps address the scarcity of forgery data, it faces several challenges. The randomly selected samples may lack the diversity and challenge needed to improve model robustness. Additionally, they can cause redundancy, increasing training time and storage requirements, and may introduce bias across users, leading to unfair training distributions. This paper proposes a novel strategy, called \textit{ProtoSig}, for generating more informative and diverse negative samples by leveraging prototypical signatures, which are compact, non-identifiable vectors obtained through a data-driven summarization of signature feature vectors. Our experiments demonstrate that ProtoSig enhances skilled forgery detection in a writer-dependent verification approach, eliminating performance variability across runs, while reducing dependence on external user data. We further demonstrate that our method achieves similar or better accuracy using a smaller dataset, offering considerable gains in scalability, computational efficiency, and storage requirements, while strengthening data privacy and promoting fairness in offline HSV systems.​

The rapid spread of information via social media has made text-based fake news detection critically important due to its societal impact.This paper presents a novel detection method called Dynamic Representation and Ensemble Selection (DRES) for identifying fake news based solely on text.DRES leverages instance hardness measures to estimate the classification difficulty for each news article across multiple textual feature representations.By dynamically selecting the textual representation and the most competent ensemble of classifiers for each instance, DRES significantly enhances prediction accuracy.Extensive experiments show that DRES achieves notable improvements over state-of-the-art methods, confirming the effectiveness of representation selection based on instance hardness and dynamic ensemble selection in boosting performance.Codes and data are available at: https://github.com/ FFarhangian/FakeNewsDetection_DRES.

Monocular facial performance capture in-the-wild is challenging due to varied capture conditions, face shapes, and expressions. Most current methods rely on linear 3D Morphable Models, which represent facial expressions independently of identity at the vertex displacement level. We propose SEREP (Semantic Expression Representation), a model that disentangles expression from identity at the semantic level. We start by learning an expression representation from high-quality 3D data of unpaired facial expressions. Then, we train a model to predict expression from monocular images relying on a novel semi-supervised scheme using low quality synthetic data. In addition, we introduce MultiREX, a benchmark addressing the lack of evaluation resources for the expression capture task. Our experiments show that SEREP outperforms state-of-the-art methods, capturing challenging expressions and transferring them to new identities.

The divide-and-conquer strategy is a common approach for solving Computer Science problems, where one divides a problem into multiple potentially simpler subproblems whose results are combined. This decomposition approach can also help solve classification problems in Machine Learning (ML). In this paper, we propose a decomposition strategy to generate multiple views from a dataset in an optimized way. The objective is to divide the input features into subsets and build ML models for each. When a new instance has to be classified, the view for which the complexity in classifying the instance is lower is chosen to be used in prediction. We show experimentally how this approach can benefit the nearest neighbor classifier, increasing classification accuracy for complex problems while overcoming the limitations of this method in handling directly high-dimensional data.

Solutions to the Algorithm Selection Problem (ASP) in machine learning face the challenge of high computational costs associated with evaluating various algorithms' performances on a given dataset. To mitigate this cost, the meta-learning field can leverage previously executed experiments shared in online repositories such as OpenML. OpenML provides an extensive collection of machine learning experiments. However, an analysis of OpenML’s records reveals limitations. It lacks diversity in pipelines, specifically when exploring data preprocessing steps/blocks, such as scaling or imputation, resulting in limited representation. Its experiments are often focused on a few popular techniques within each pipeline block, leading to an imbalanced sample. To overcome the observed limitations of OpenML, we propose PIPES, a collection of experiments involving multiple pipelines designed to represent all combinations of the selected sets of techniques, aiming at diversity and completeness. PIPES stores the results of experiments performed applying 9,408 pipelines to 300 datasets. It includes detailed information on the pipeline blocks, training and testing times, predictions, performances, and the eventual error messages. This comprehensive collection of results allows researchers to perform analyses across diverse and representative pipelines and datasets. PIPES also offers potential for expansion, as additional data and experiments can be incorporated to support the meta-learning community further. The data, code, supplementary material, and all experiments can be found at https://github.com/cynthiamaia/PIPES.git.

Psychological biases, such as confirmation bias, make individuals particularly vulnerable to believing and spreading fake news on social media, leading to significant consequences in domains such as public health and politics. Machine learning–based fact-checking systems have been widely studied to mitigate this problem. Among them, ensemble methods are particularly effective in combining multiple classifiers to improve robustness. However, their performance heavily depends on the diversity of the constituent classifiers—selecting genuinely diverse models remains a key challenge, especially when models tend to learn redundant patterns. In this work, we propose a novel automatic classifier selection approach that prioritizes diversity, also extended by performance. The method first computes pairwise diversity between classifiers and applies hierarchical clustering to organize them into groups at different levels of granularity. A HierarchySelect then explores these hierarchical levels to select one pool of classifiers per level, each representing a distinct intra-pool diversity. The most diverse pool is identified and selected for ensemble construction from these. The selection process incorporates an evaluation metric reflecting each classifier’s performance to ensure the ensemble also generalises well. We conduct experiments with 40 heterogeneous classifiers across six datasets from different application domains and with varying numbers of classes. Our method is compared against the Elbow heuristic and state-of-the-art baselines. Results show that our approach achieves the highest accuracy on two of six datasets. The implementation details are available on the project’s repository: https://github.com/SaraBCoutinho/HSFN.

A key challenge in visible-infrared person re-identification (V-I ReID) is training a backbone model capable of effectively addressing the significant discrepancies across modalities. State-of-the-art methods that generate a single intermediate bridging domain are often less effective, as this generated domain may not adequately capture sufficient common discriminant information. This paper introduces Bidirectional Multi-step Domain Generalization (BMDG), a novel approach for unifying feature representations across diverse modalities. BMDG creates multiple virtual intermediate domains by learning and aligning body part features extracted from both I and V modalities. In particular, our method aims to minimize the cross-modal gap in two steps. First, BMDG aligns modalities in the feature space by learning shared and modality-invariant body part prototypes from V and I images. Then, it generalizes the feature representation by applying bidirectional multi-step learning, which progressively refines feature representations in each step and incorporates more prototypes from both modalities. Based on these prototypes, multiple bridging steps enhance the feature representation. Experiments11Our code is available at: alehdaghi.github.io/BMDG conducted on V-I ReID datasets indicate that our BMDG approach can outperform state-of-the-art part-based and intermediate generation methods, and can be integrated into other part-based methods to enhance their V-I ReID performance.

Visible-infrared person re-identification (V-I ReID) seeks to retrieve images of the same individual captured over a distributed network of RGB and IR sensors. Several V-I ReID approaches directly integrate the V and I modalities to represent images within a shared space. However, given the significant gap in the data distributions between V and I modalities, cross-modal V-I ReID remains challenging. A solution is to involve a privileged intermediate space to bridge between modalities, but in practice, such data is not available and requires selecting or creating effective mechanisms for informative intermediate domains. This paper introduces the Adaptive Generation of Privileged Intermediate Information (AGPI2) training approach to adapt and generate a virtual domain that bridges discriminative information between the V and I modalities. AGPI2enhances the training of a deep V-I ReID backbone by generating and then leveraging bridging privileged information without modifying the model in the inference phase. This information captures shared discriminative attributes that are not easily ascertainable for the model within individual V or I modalities. Towards this goal, a non-linear generative module is trained with adversarial objectives, transforming V attributes into intermediate spaces that also contain I features. This domain exhibits less domain shift relative to the I domain compared to the V domain. Meanwhile, the embedding module within AGPI2aims to extract discriminative modality-invariant features for both modalities by leveraging modality-free descriptors from generated images, making them a bridge between the main modalities. Experiments conducted on challenging V-I ReID datasets indicate that AGPI2consistently increases matching accuracy without additional computational resources during inference.

Dataset scaling, a.k.a. normalization, is an essential preprocessing step in a machine learning (ML) pipeline. It aims to adjust the scale of attributes in a way that they all vary within the same range. This transformation is known to improve the performance of classification models. Still, there are several scaling techniques (STs) to choose from, and no ST is guaranteed to be the best for a dataset regardless of the classifier chosen. It is thus a problem- and classifier-dependent decision. Furthermore, there can be a huge difference in performance when selecting the wrong technique; hence, it should not be neglected. That said, the trial-and-error process of finding the most suitable technique for a particular dataset can be unfeasible. As an alternative, we propose the Meta-scaler, which uses meta-learning (MtL) to build meta-models to automatically select the best ST for a given dataset and classification algorithm. The meta-models learn to represent the relationship between meta-features extracted from the datasets and the performance of specific classification algorithms on these datasets when scaled with different techniques. Our experiments using 12 base classifiers, 300 datasets, and five STs demonstrate the feasibility and effectiveness of the approach. When using the ST selected by the Meta-scaler for each dataset, 10 of 12 base models tested achieved statistically significantly better classification performance than any fixed choice of a single ST. The Meta-scaler also outperforms state-of-the-art MtL approaches for ST selection. The source code, data, and results from the experiments in this article are available at a GitHub repository (https://github.com/amorimlb/meta_scaler).

Reconstructing an avatar from a portrait image has many applications in multimedia, but remains a challenging research problem. Extracting reflectance maps and geom- etry from one image is ill-posed: recovering geometry is a one-to-many mapping problem and reflectance and light are difficult to disentangle. Accurate geometry and reflectance can be captured under the controlled conditions of a light stage, but it is costly to acquire large datasets in this fash- ion. Moreover, training solely with this type of data leads to poor generalization with in-the-wild images. This moti- vates the introduction of MoSAR, a method for 3D avatar generation from monocular images. We propose a semi- supervised training scheme that improves generalization by learning from both light stage and in-the-wild datasets. This is achieved using a novel differentiable shading formulation. We show that our approach effectively disentangles the intrinsic face parameters, producing relightable avatars. As a result, MoSAR11Project page: https://ubisoft-laforge.github.io/character/mosar estimates a richer set of skin reflectance maps and generates more realistic avatars than existing state-of-the-art methods. We also release a new dataset, that provides intrinsic face attributes (diffuse, specular, am- bient occlusion and translucency maps) for 10k subjects.

Dynamic Selection (DS), where base classifiers are chosen from a classifier’s pool for each new instance at test time, has shown to be highly effective in pattern recognition. However, instability and redundancy in the classifier pools can impede computational efficiency and accuracy in dynamic ensemble selection. This paper introduces a meta-learning recommendation system (MLRS) to recommend the optimal pool generation scheme for DES methods tailored to individual datasets. The system employs a meta-model built from dataset meta-features to predict the most suitable pool generation scheme and DES method for a given dataset. Through an extensive experimental study encompassing 288 datasets, we demonstrate that this meta-learning recommendation system outperforms traditional fixed pool or DES method selection strategies, highlighting the efficacy of a meta-learning approach in refining DES method selection. The source code, datasets, and supplementary results can be found in this project’s GitHub repository: https://github.com/Menelau/MLRS-PDS.

Data normalization is an essential task when modeling a classification system. When dealing with data streams, data normalization becomes especially challenging since we may not know in advance the properties of the features, such as their minimum/maximum values, and these properties may change over time. We compare the accuracies generated by eight well-known distance functions in data streams without normalization, normalized considering the statistics of the first batch of data received, and considering the previous batch received. We argue that experimental protocols for streams that consider the full stream as normalized are unrealistic and can lead to biased and poor results. Our results indicate that using the original data stream without applying normalization, and the Canberra distance, can be a good combination when no information about the data stream is known beforehand.

Dynamic Ensemble Selection (DES) is a Multiple Classifier Systems (MCS) approach that aims to select an ensemble for each query sample during the selection phase. Even with the proposal of several DES approaches, no particular DES technique is the best choice for different problems. Thus, we hypothesize that selecting the best DES approach per query instance can lead to better accuracy. To evaluate this idea, we introduce the Post-Selection Dynamic Ensemble Selection (PS-DES) approach, a post-selection scheme that evaluates ensembles selected by several DES techniques using different metrics. Experimental results show that using accuracy as a metric to select the ensembles, PS-DES performs better than individual DES techniques. PS-DES source code is available in a GitHub repository44https://github.com/prgc/ps-des.

The Android operating system provides functions and methods to handle sensitive data to secure users’ data. The Android security literature extracts binary features from a method and classifies the method into one of the Security Relevant Method's classes, adding information about how the method handles sensitive data. However, the usage of binary features hinders the performance of some classifiers due to the high collision rate between instances. Although previous works have explored Security Relevant Method classification, an extensive study of machine learning algorithms over this problem has not been conceived. This work fills this gap, analyzing Monolithic classifiers, Multiple Classifier Systems, and Embedding algorithms to transform binary features into real-valued features, aiming to facilitate the classifier's work by minimizing the ambiguity promoted by the collision. Our analyzes show that META-DES, using a pool of Decision Trees trained with the Random Forest algorithm, statistically has the best results. We also find that, in general, distance-based classifiers have a disadvantage in binary features. Moreover, embedding techniques such as deep metric learning with triplet loss can reduce geometrical instance ambiguity, improving the performance of the weakest learning algorithms. However, its usage was detrimental to the performance of more robust techniques, such as dynamic ensemble models better suited for handling difficult cases. The dataset and code used for the experiments are available in the following repository:https://github.com/walbermr/android-srm-ml-evaluation.

Despite the recent success of deep learning architectures, person re-identification (ReID) remains a challenging problem in real-word applications. Several unsupervised single-target domain adaptation (STDA) methods have recently been proposed to limit the decline in ReID accuracy caused by the domain shift that typically occurs between source and target video data. Given the multimodal nature of person ReID data (due to variations across camera viewpoints and capture conditions), training a common CNN backbone to address domain shifts across multiple target domains, can provide an efficient solution for real-time ReID applications. Although multi-target domain adaptation (MTDA) has not been widely addressed in the ReID literature, a straightforward approach consists in blending different target datasets, and performing STDA on the mixture to train a common CNN. However, this approach may lead to poor generalization, especially when blending a growing number of distinct target domains to train a smaller CNN. To alleviate this problem, we introduce a new MTDA method based on knowledge distillation (KD-ReID) that is suitable for real-time person ReID applications. Our method adapts a common lightweight student backbone CNN over the target domains by alternatively distilling from multiple specialized teacher CNNs, each one adapted on data from a specific target domain. Extensive experiments1conducted on several challenging person ReID datasets indicate that our approach outperforms state-of-art methods for MTDA, including blending methods, particularly when training a compact CNN backbone like OSNet. Results suggest that our flexible MTDA approach can be employed to design cost-effective ReID systems for real-time video surveillance applications.

Most dynamic ensemble selection (DES) methods utilize the K-Nearest Neighbors (KNN) algorithm to estimate the competence of classifiers in a small region surrounding the query sample. However, KNN is very sensitive to the local distribution of the data. Moreover, it also has a high computational cost as it requires storing the whole data in memory and performing multiple distance calculations during inference. Hence, the de-pendency on the KNN algorithm ends up limiting the use of DES techniques for large-scale problems. This paper presents a new DES framework based on fuzzy hyperboxes called FH-DES. Each hyperbox can represent a group of samples using only two data points (Min and Max corners). Thus, the hyperbox-based system will have less computational complexity than other dynamic se-lection methods. In addition, despite the KNN-based approaches, the fuzzy hyperbox is not sensitive to the local data distribution. Therefore, the local distribution of the samples does not affect the system's performance. Furthermore, in this research, for the first time, misclassified samples are used to estimate the competence of the classifiers, which has not been observed in previous fusion approaches. Experimental results demonstrate that the proposed method has high classification accuracy while having a lower complexity when compared with the state-of-the-art dynamic selection methods. The implemented code is available at https://github.com/redavtalab/FH-DES_IJCNN.git.

Deep Siamese trackers have recently gained much attention in recent years since they can track visual objects at high speed. Additionally, adaptive tracking methods, where target samples collected by the tracker are employed for online learning, have achieved state-of-the-art accuracy. However, single object tracking (SOT) remains a challenging task in real-world application due to changes and deformations in a target object's appearance. Learning on all the collected samples may lead to catastrophic forgetting, and thereby corrupt the tracking model. In this paper, SOT is formulated as an online incremental learning problem. A new method is proposed for dynamic sample selection and memory replay, preventing template corruption. In particular, we propose a change detection mechanism to detect gradual changes in object appearance, and select the corresponding samples for online adaption. In addition, an entropy-based sample selection strategy is introduced to maintain a diversified auxiliary buffer for memory replay. Our proposed method can be integrated into any object tracking algorithm that leverages online learning for model adaptation. Extensive experiments conducted on the OTB-100, LaSOT, UAV123, and TrackingNet datasets highlight the cost-effectiveness of our method, along with the contribution of its key components. Results indicate that integrating our proposed method into state-of-art adaptive Siamese trackers can increase the potential benefits of a template update strategy, and significantly improve performance. Code: https://github.com/madhukiranets/Adaptive-Siamese-Dimp

Class imbalance is a characteristic known for making learning more challenging for classification models as they may end up biased towards the majority class. A promising approach among the ensemble-based methods in the context of imbalance learning is Dynamic Selection (DS). DS techniques single out a subset of the classifiers in the ensemble to label each given unknown sample according to their estimated competence in the area surrounding the query. Because only a small region is taken into account in the selection scheme, the global class disproportion may have less impact over the system's performance. However, the presence of local class overlap may severely hinder the DS techniques' performance over imbalanced distributions as it not only exacerbates the effects of the under-representation but also introduces ambiguous and possibly unreliable samples to the competence estimation process. Thus, in this work, we propose a DS technique which attempts to minimize the effects of the local class overlap during the classifier selection procedure. The proposed method iteratively removes from the target region the instance perceived as the hardest to classify until a classifier is deemed competent to label the query sample. The known samples are characterized using instance hardness measures that quantify the local class overlap. Experimental results show that the proposed technique can significantly outperform the baseline as well as several other DS techniques, suggesting its suitability for dealing with class under-representation and overlap. Furthermore, the proposed technique still yielded competitive results when using an under-sampled, less overlapped version of the labelled sets, specially over the problems with a high proportion of minority class samples in overlap areas. Code available at https://github.com/marianaasouza/lords.

In writer-independent verification systems, a single model is trained for all users of the system using dissimilarity vectors obtained through a dichotomy transformation that converts a multi-class problem into a 2-class problem comprising: (i) the intra-class dissimilarity vectors computed from samples of the same user, (ii) the inter-class dissimilarity vectors computed from samples of different users. When mapping handwritten signature representations, it is desired to obtain well-separated dense clusters of signature representations for each user, in such a way that transformed intra-class dissimilarity vectors tend to be separated from the inter-class dissimilarity vectors. Moreover, since skilled forgeries resemble reference signatures, it is also desired to obtain skilled forgery dissimilarity vectors that are further away from the region of the intra-class dissimilarity vectors. In this work, it is hypothesized that an improved dissimilarity space can be achieved through a multi-task framework for learning handwritten signature feature representations based on deep contrastive learning. The proposed framework is composed of two objective-specific tasks; it does not use skilled forgeries for training. The first task aims to map signature examples of a given user in a dense cluster, while linearly separating the signature representations of different users. The second task aims to adjust forgery representations by adopting a contrastive loss with the ability to perform hard negative mining. Hard negatives are similar examples but from different classes that can be seen as artificially generated skilled forgeries for training. In a writer-independent verification approach, the model obtained with the proposed framework is evaluated in terms of the equal error rate on GPDS-300, CEDAR and MCYT-75 datasets. Experiments demonstrated a statistically significant improvement in signature verification compared to the state-of-the-art SigNet feature extraction method.

SigNet is a state of the art model for feature representation used for handwritten signature verification (HSV). This representation is based on a Deep Convolutional Neural Network (DCNN) and contains 2048 dimensions. When transposed to a dissimilarity space generated by the dichotomy transformation (DT), related to the writer-independent (WI) approach, these features may include redundant information. This paper investigates the presence of overfitting when using Binary Particle Swarm Optimization (BPSO) to perform the feature selection in a wrapper mode. We proposed a method based on a global validation strategy with an external archive to control overfitting during the search for the most discriminant representation. Moreover, an investigation is also carried out to evaluate the use of the selected features in a transfer learning context. The analysis is carried out on a writer-independent approach on the CEDAR, MCYT and GPDS datasets. The experimental results showed the presence of overfitting when no validation is used during the optimization process and the improvement when the global validation strategy with an external archive is used. Also, the space generated after feature selection can be used in a transfer learning context.

Dynamic selection techniques aim at selecting the local experts around each test sample in particular for performing its classification. While generating the classifier on a local scope may make it easier for singling out the locally competent ones, as in the online local pool (OLP) technique, using the same base-classifier model in uneven distributions may restrict the local level of competence, since each region may have a data distribution that favors one model over the others. Thus, we propose in this work a problem-independent dynamic base-classifier model recommendation for the OLP technique, which uses information regarding the behavior of a portfolio of models over the samples of different problems to recommend one (or several) of them in a per-instance manner. Our proposed framework builds a multi-label meta-classifier responsible for recommending a set of relevant base-classifier models based on the local data complexity of the region surrounding each test sample. The OLP technique then produces a local pool with the model that yields the highest probability score of the meta-classifier. Experimental results show that different data distributions favored different model types on a local scope. Moreover, based on the performance of an ideal model type selector, it was observed that there is a clear advantage in choosing a relevant base-classifier model for each test instance in particular. Overall, the proposed model type recommender system yielded a statistically similar performance to the original OLP with fixed base-classifier model. However, the proposed framework struggled to recommend at least one relevant model type specially for the samples with low labelset cardinality. Given the novelty of the approach and the gap in performance between the proposed framework and the ideal selector, we regard this as a promising research direction. Code available at github.com/marianaasouza/dynamic-model-recommender.

DESlib is an open-source python library providing the implementation of several dynamic selection techniques. The library is divided into three modules: (i) dcs, containing the implementation of dynamic classifier selection methods (DCS); (ii) des, containing the implementation of dynamic ensemble selection methods (DES); (iii) static, with the implementation of static ensemble techniques. The library is fully documented (documentation available online on Read the Docs), has a high test coverage (codecov.io) and is part of the scikit-learn-contrib supported projects. Documentation, code and examples can be found on its GitHub page: https://github.com/scikit-learn-contrib/DESlib.

Imbalanced problems are characterized by a disproportion between the number of samples from the classes in a classification problem. This difference in amount of examples may lead to a bias toward the majority class, hindering the recognition of the underrepresented minority class. Ensemble methods have been widely used for dealing with such problems, and have been shown to perform well on them. In this context, Dynamic Selection (DS) approaches, which perform the classification task on a local level, have been receiving some attention for their promising results. More specifically, the Frienemy Indecision Region Dynamic Ensemble Selection++ (FIRE-DES++) framework, which has yielded state-of-the-art results on imbalanced problems, use a data preprocessing technique for noise removal and a class-balanced neighborhood definition for coping with imbalanced datasets. A different DS-based approach proposed in a previous work, an online local pool generation method, generates on the fly locally accurate classifiers for labelling samples near the class borders. Though the local generation of the classifiers may reduce the impact of class imbalance on the performance of the technique, its suitability for imbalance learning was not yet evaluated. Thus, in this work we evaluate how well the online local pool generation method deals with imbalanced problems. We perform a comparative analysis with a baseline technique using three Dynamic Classifier Selection (DCS) techniques over 64 imbalanced datasets and four performance measures. We also evaluate the use of the preprocessing and balanced neighborhood definition steps from the FIRE-DES++ on the online scheme to assess their impact on the performance of the method. Moreover, we evaluate the online technique and its variants against seven state-of-the-art ensemble methods, including both static and DS approaches. Experimental results show that the approach of locally generating the classifiers is advantageous for imbalance learning, providing an improvement to the DCS techniques and yielding state-of-the-art results. Furthermore, the addition of the noise removal and the balanced neighborhood definition steps to the online scheme improved the overall results of the technique, which indicates the advantage of including such steps in DS-based techniques.

When compared to Writer-Dependent (WD) Handwritten Signature Verification, in which a model is trained for each individual writer, the Writer-Independent (WI) approach offers greater scalability, since only a single model is trained for all users from a dissimilarity space generated by the dichotomy transformation. However, many samples from the dissimilarity space are redundant and have little influence during the training of the verification model. This work investigates whether prototype selection (PS) preprocessing can be used in the space resulting from the dichotomy transformation without degrading the performance of the classifier. Furthermore, an investigation is also performed to examine the use of a WI classifier in a transfer learning scenario, i.e., where the classifier is trained in one dataset, and is used to verify signatures in other datasets. The experiments reported herein show that the use of prototype selection in the dissimilarity space allows a reduction in the complexity of the classifier without degrading its generalization performance. In addition, the results show that the WI classifier is scalable enough to be used in a transfer learning approach, with a resulting performance comparable to that of a classifier trained and tested in the same dataset. An analysis of the results obtained based on the instance hardness (IH) measure and dendrogram diagrams is performed in order to better understand the behavior of the resulting dichotomy transformation.

Class imbalance refers to classification problems in which many more instances are available for certain classes than for others. Such imbalanced datasets require special attention because traditional classifiers generally favor the majority class which has a large number of instances. Ensemble of classifiers has been reported to yield promising results. However, the majority of ensemble methods applied to imbalance learning are static ones. Moreover, they only deal with binary imbalanced problems. Hence, this paper presents an empirical analysis of Dynamic Selection techniques and data preprocessing methods for dealing with multi-class imbalanced problems. We considered five variations of preprocessing methods and 14 Dynamic Selection schemes. Our experiments conducted on 26 multi-class imbalanced problems show that the dynamic ensemble improves the AUC and the [Formula: see text]-mean as compared to the static ensemble. Moreover, data preprocessing plays an important role in such cases.

Dynamic Ensemble Selection (DES) techniques aim to select locally competent classifiers for the classification of each new test sample. Most DES techniques estimate the competence of classifiers using a given criterion over the region of competence of the test sample (its the nearest neighbors in the validation set). The K-Nearest Oracles Eliminate (KNORA-E) DES selects all classifiers that correctly classify all samples in the region of competence of the test sample, if such classifier exists, otherwise, it removes from the region of competence the sample that is furthest from the test sample, and the process repeats. When the region of competence has samples of different classes, KNORAE can reduce the region of competence in such a way that only samples of a single class remain in the region of competence, leading to the selection of locally incompetent classifiers that classify all samples in the region of competence as being from the same class. In this paper, we propose two DES techniques: KNearest Oracles Borderline (KNORA-B) and K-Nearest Oracles Borderline Imbalanced (KNORA-BI). KNORA-B is a DES technique based on KNORA-E that reduces the region of competence but maintains at least one sample from each class that is in the original region of competence. KNORA-BI is a variation of KNORA-B for imbalance datasets that reduces the region of competence but maintains at least one minority class sample if there is any in the original region of competence. Experiments are conducted comparing the proposed techniques with 19 DES techniques from the literature using 40 datasets. The results show that the proposed techniques achieved interesting results, with KNORA-BI outperforming state-of-art techniques.

In Machine Learning, ensemble methods have been receiving a great deal of attention. Techniques such as Bagging and Boosting have been successfully applied to a variety of problems. Nevertheless, such techniques are still susceptible to the effects of noise and outliers in the training data. We propose a new method for the generation of pools of classifiers based on Bagging, in which the probability of an instance being selected during the resampling process is inversely proportional to its instance hardness, which can be understood as the likelihood of an instance being misclassified, regardless of the choice of classifier. The goal of the proposed method is to remove noisy data without sacrificing the hard instances which are likely to be found on class boundaries. We evaluate the performance of the method in nineteen public data sets, and compare it to the performance of the Bagging and Random Subspace algorithms. Our experiments show that in high noise scenarios the accuracy of our method is significantly better than that of Bagging.

In dynamic selection (DS) techniques, only the most competent classifiers, for the classification of a specific test sample are selected to predict the sample's class labels. The more important step in DES techniques is estimating the competence of the base classifiers for the classification of each specific test sample. The classifiers' competence is usually estimated using the neighborhood of the test sample defined on the validation samples, called the region of competence. Thus, the performance of DS techniques is sensitive to the distribution of the validation set. In this paper, we evaluate six prototype selection techniques that work by editing the validation data in order to remove noise and redundant instances. Experiments conducted using several state-of-the-art DS techniques over 30 classification problems demonstrate that by using prototype selection techniques we can improve the classification accuracy of DS techniques and also significantly reduce the computational cost involved.

The Oracle model has been used not only for comparison between techniques but also in the design of different methods in Multiple Classifier Systems (MCS). Even though the model represents the ideal classifier selection scheme, Dynamic Classifier Selection (DCS) techniques present a large performance gap from the Oracle. This means that, for a significant number of instances, the DCS techniques are not able to select a competent classifier, despite the Oracles assurance of its presence in the pool. Given that issue, this work aims to investigate the reasons why the Oracle model may not be well suited for guiding the search for a promising pool of classifiers for DCS techniques. For this purpose, a pool generation method that guarantees an Oracle accuracy rate of 100% in the training set is proposed. This method is further used to analyse the behavior of DCS techniques when the presence of at least one competent classifier in the pool for each training sample is assured. Experiments show that integrating Oracle information in the generation phase of an MCS has little impact on the gap between the accuracy rates of DCS techniques and the Oracle. Moreover, it is also shown that, for a theoretical limit of 100%, the DCS techniques were only able to select a competent classifier for at most 85% of the instances, on average. Results suggest that the Oracle is not the best guide for generating a pool of classifiers for DCS, for the model is performed globally whilst DCS techniques work with local data only.

Dynamic selection (DS) is a mechanism to select one or an ensemble of competent classifiers from a pool of base classifiers, in order to classify a specific test sample. The size of this pool is user defined and yet crucial to control the computational complexity and performance of a DS. An appropriate pool size depends on the choice of base classifiers, the underlying DS method used, and more importantly, the characteristics of the given problem. After the DS method and the base classifiers are selected, an appropriate pool size for a given problem can be obtained by the repetitive application of the DS with a variety of sizes, after which a selection is performed. Since this brute force approach is computationally expensive, researchers usually set the pool size to a pre-specified value. However, this strategy may reduce the performance of the DS method. Instead, we propose a meta-regression model in order to predict a suitable pool size, based on the intrinsic classification complexity of a problem. In our strategy, we obtain the best pool sizes for a number of data sets, using the brute force approach. Additionally, we extract meta-features that represent classification complexity of a problem. These two pieces of information are associated by means of meta-regression models. Finally, for an unseen problem, we predict the pool size using this model and the classification complexity information.We carry out the experiments on 64 two-class data sets and with several well-known DS methods. We also consider variants of meta-regression techniques and report prediction results. We further analyze these results using a statistical test. Finally, we investigate the performance of a DS and observe that DS performs equivalently for predicted and the best pool sizes.

In Dynamic Ensemble Selection (DES) techniques, only the most competent classifiers are selected to classify a given query sample. Hence, the key issue in DES is how to estimate the competence of each classifier in a pool to select the most competent ones. In order to deal with this issue, we proposed a novel dynamic ensemble selection framework using meta-learning, called META-DES. The framework is divided into three steps. In the first step, the pool of classifiers is generated from the training data. In the second phase the meta-features are computed using the training data and used to train a meta-classifier that is able to predict whether or not a base classifier from the pool is competent enough to classify an input instance. In this paper, we propose improvements to the training and generalization phase of the META-DES framework. In the training phase, we evaluate four different algorithms for the training of the meta-classifier. For the generalization phase, three combination approaches are evaluated: Dynamic selection, where only the classifiers that attain a certain competence level are selected; Dynamic weighting, where the meta-classifier estimates the competence of each classifier in the pool, and the outputs of all classifiers in the pool are weighted based on their level of competence; and a hybrid approach, in which first an ensemble with the most competent classifiers is selected, after which the weights of the selected classifiers are estimated in order to be used in a weighted majority voting scheme. Experiments are carried out on 30 classification datasets. Experimental results demonstrate that the changes proposed in this paper significantly improve the recognition accuracy of the system in several datasets.

In this paper, we propose a novel dynamic ensemble selection framework using meta-learning. The framework is divided into three steps. In the first step, the pool of classifiers is generated from the training data. The second phase is responsible to extract the meta-features and train the meta-classifier. Five distinct sets of meta-features are proposed, each one corresponding to a different criterion to measure the level of competence of a classifier for the classification of a given query sample. The meta-features are computed using the training data and used to train a meta-classifier that is able to predict whether or not a base classifier from the pool is competent enough to classify an input instance. Three different training scenarios for the training of the meta-classifier are considered: problem-dependent, problem-independent and hybrid. Experimental results show that the problem-dependent scenario provides the best result. In addition, the performance of the problem-dependent scenario is strongly correlated with the recognition rate of the system. A comparison with state-of-the-art techniques shows that the proposed-dependent approach outperforms current dynamic ensemble selection techniques.

Dynamic classifier selection systems aim to select a group of classifiers that is most adequate for a specific query pattern. This is done by defining a region around the query pattern and analyzing the competence of the classifiers in this region. However, the regions are often surrounded by noise which can difficult the classifier selection. This fact makes the performance of most dynamic selection systems no better than static selections. In this paper we demonstrate that the performance of dynamic selection systems end up limited by the quality of the regions extracted. Thereafter, we propose a new dynamic classifier selection system that improves the regions of competence in order to achieve higher recognition rates. Results obtained from several classification databased show the proposed method not only significantly increase the recognition performance, but also decreases the computational cost.

This paper presents a novel approach for cursive character recognition by using multiple feature extraction algorithms and a classifier ensemble. Several feature extraction techniques, using different approaches, are extracted and evaluated. Two techniques, Modified Edge Maps and Multi Zoning, are proposed. The former one presents the best overall result. Based on the results, a combination of the feature sets is proposed in order to achieve high recognition performance. This combination is motivated by the observation that the feature sets are both, independent and complementary. The ensemble is performed by combining the outputs generated by the classifier in each feature set separately. Both fixed and trained combination rules are evaluated using the C-Cube database. A trained combination scheme using a MLP network as combiner achieves the best results which is also the best results for the C-Cube database by a good margin.