Search

In recent times, the progress of machine learning has facilitated the development of decision support systems that exhibit predictive accuracy, surpassing human capabilities in certain scenarios. However, this improvement has come at the cost of increased model complexity, rendering them black-box models that obscure their internal logic from users. These black boxes are primarily designed to optimize predictive accuracy, limiting their applicability in critical domains such as medicine, law, and finance, where both accuracy and interpretability are crucial factors for model acceptance. Despite the growing body of research on interpretability, there remains a significant dearth of evaluation methods for the proposed approaches. This survey aims to shed light on various evaluation methods employed in interpreting models. Two primary procedures are prevalent in the literature: qualitative and quantitative evaluations. Qualitative evaluations rely on human assessments, while quantitative evaluations utilize computational metrics. Human evaluation commonly manifests as either researcher intuition or well-designed experiments. However, this approach is susceptible to human biases and fatigue and cannot adequately compare two models. Consequently, there has been a recent decline in the use of human evaluation, with computational metrics gaining prominence as a more rigorous method for comparing and assessing different approaches. These metrics are designed to serve specific goals, such as fidelity, comprehensibility, or stability. The existing metrics often face challenges when scaling or being applied to different types of model outputs and alternative approaches. Another important factor that needs to be addressed is that while evaluating interpretability methods, their results may not always be entirely accurate. For instance, relying on the drop in probability to assess fidelity can be problematic, particularly when facing the challenge of out-of-distribution data. Furthermore, a fundamental challenge in the interpretability domain is the lack of consensus regarding its definition and requirements. This issue is compounded in the evaluation process and becomes particularly apparent when assessing comprehensibility. [ABSTRACT FROM AUTHOR]

Sentiment is currently one of the most emerging areas of research due to the large amount of web content coming from social networking websites. Sentiment analysis is a crucial process for recommending systems for most people. Generally, the purpose of sentiment analysis is to determine an author's attitude toward a subject or the overall tone of a document. There is a huge collection of studies that make an effort to predict how useful online reviews will be and have produced conflicting results on the efficacy of different methodologies. Furthermore, many of the current solutions employ manual feature generation and conventional shallow learning methods, which restrict generalization. As a result, the goal of this research is to develop a general approach using transfer learning by applying the "BERT (Bidirectional Encoder Representations from Transformers)"-based model. The efficiency of BERT classification is then evaluated by comparing it with similar machine learning techniques. In the experimental evaluation, the proposed model demonstrated superior performance in terms of outstanding prediction and high accuracy compared to earlier research. Comparative tests conducted on positive and negative Yelp reviews reveal that fine-tuned BERT classification performs better than other approaches. In addition, it is observed that BERT classifiers using batch size and sequence length significantly affect classification performance. [ABSTRACT FROM AUTHOR]

Understanding the reasoning behind a predictive model's decision is an important and longstanding problem driven by ethical and legal considerations. Most recent research has focused on the interpretability of supervised models, whereas unsupervised learning has received less attention. However, the majority of the focus was on interpreting the whole model in a manner that undermined accuracy or model assumptions, while local interpretation received much less attention. Therefore, we propose an intrinsic interpretation for the Gaussian mixture model that provides both global insight and local interpretations. We employed the Bhattacharyya coefficient to measure the overlap and divergence across clusters to provide a global interpretation in terms of the differences and similarities between the clusters. By analyzing the GMM exponent with the Garthwaite–Kock corr-max transformation, the local interpretation is provided in terms of the relative contribution of each feature to the overall distance. Experimental results obtained on three datasets show that the proposed interpretation method outperforms the post hoc model-agnostic LIME in determining the feature contribution to the cluster assignment. [ABSTRACT FROM AUTHOR]

Boolean structural measures were introduced to explain the high performance of conflict-driven clause-learning (CDCL) SAT solvers on industrial SAT instances. Those considered in this study include measures related to backbones and backdoors: backbone size, backbone frequency, and backdoor size. A key area of research is to improve the performance of CDCL SAT solvers by exploiting these measures. For the purpose of guiding the CDCL SAT solver for branching on backbone and backdoor variables, this study proposes low-overhead heuristics for computing these variables. Through these heuristics, a set of modifications to the Variable State Independent Decaying Sum (VSIDS) decision heuristic is suggested to exploit backbones and backdoors and potentially improve the performance of CDCL SAT solvers. In total, fifteen variants of two competitive base solvers, MapleLCMDistChronoBT-DL-v3 and LSTech, were developed. Empirical evaluation was conducted on 32 industrial families from 2002–2021 SAT competitions. According to the results, modifying the VSIDS heuristic in the base solvers to exploit backbones and backdoors improves its performance. In particular, our new CDCL SAT solver, LSTech_BBsfcr_v1, solved more industrial SAT instances than the winning CDCL SAT solvers in 2020 and 2021 SAT competitions. [ABSTRACT FROM AUTHOR]

A high-performance versatile computer-assisted pronunciation training (CAPT) system that provides the learner immediate feedback as to whether their pronunciation is correct is very helpful in learning correct pronunciation and allows learners to practice this at any time and with unlimited repetitions, without the presence of an instructor. In this paper, we propose deep learning-based techniques to build a high-performance versatile CAPT system for mispronunciation detection and diagnosis (MDD) and articulatory feedback generation for non-native Arabic learners. The proposed system can locate the error in pronunciation, recognize the mispronounced phonemes, and detect the corresponding articulatory features (AFs), not only in words but even in sentences. We formulate the recognition of phonemes and corresponding AFs as a multi-label object recognition problem, where the objects are the phonemes and their AFs in a spectral image. Moreover, we investigate the use of cutting-edge neural text-to-speech (TTS) technology to generate a new corpus of high-quality speech from predefined text that has the most common substitution errors among Arabic learners. The proposed model and its various enhanced versions achieved excellent results. We compared the performance of the different proposed models with the state-of-the-art end-to-end technique of MDD, and our system had a better performance. In addition, we proposed using fusion between the proposed model and the end-to-end model and obtained a better performance. Our best model achieved a 3.83% phoneme error rate (PER) in the phoneme recognition task, a 70.53% F1-score in the MDD task, and a detection error rate (DER) of 2.6% for the AF detection task. [ABSTRACT FROM AUTHOR]

Sign language is the main channel for hearing-impaired people to communicate with others. It is a visual language that conveys highly structured components of manual and non-manual parameters such that it needs a lot of effort to master by hearing people. Sign language recognition aims to facilitate this mastering difficulty and bridge the communication gap between hearing-impaired people and others. This study presents an efficient architecture for sign language recognition based on a convolutional graph neural network (GCN). The presented architecture consists of a few separable 3DGCN layers, which are enhanced by a spatial attention mechanism. The limited number of layers in the proposed architecture enables it to avoid the common over-smoothing problem in deep graph neural networks. Furthermore, the attention mechanism enhances the spatial context representation of the gestures. The proposed architecture is evaluated on different datasets and shows outstanding results. [ABSTRACT FROM AUTHOR]

This study proposes using object detection techniques to recognize sequences of articulatory features (AFs) from speech utterances by treating AFs of phonemes as multi-label objects in speech spectrogram. The proposed system, called AFD-Obj, recognizes sequence of multi-label AFs in speech signal and localizes them. AFD-Obj consists of two main stages: firstly, we formulate the problem of AFs detection as an object detection problem and prepare the data to fulfill requirement of object detectors by generating a spectral three-channel image from the speech signal and creating the corresponding annotation for each utterance. Secondly, we use annotated images to train the proposed system to detect sequences of AFs and their boundaries. We test the system by feeding spectrogram images to the system, which will recognize and localize multi-label AFs. We investigated using these AFs to detect the utterance phonemes. YOLOv3-tiny detector is selected because of its real-time property and its support for multi-label detection. We test our AFD-Obj system on Arabic and English languages using KAPD and TIMIT corpora, respectively. Additionally, we propose using YOLOv3-tiny as an Arabic phoneme detection system (i.e., PD-Obj) to recognize and localize a sequence of Arabic phonemes from whole speech utterances. The proposed AFD-Obj and PD-Obj systems achieve excellent results for Arabic corpus and comparable to the state-of-the-art method for English corpus. Moreover, we showed that using only one-scale detection is suitable for AFs detection or phoneme recognition. [ABSTRACT FROM AUTHOR]

Nowadays, Internet of Things (IoT) technology has various network applications and has attracted the interest of many research and industrial communities. Particularly, the number of vulnerable or unprotected IoT devices has drastically increased, along with the amount of suspicious activity, such as IoT botnet and large-scale cyber-attacks. In order to address this security issue, researchers have deployed machine and deep learning methods to detect attacks targeting compromised IoT devices. Despite these efforts, developing an efficient and effective attack detection approach for resource-constrained IoT devices remains a challenging task for the security research community. In this paper, we propose an efficient and effective IoT botnet attack detection approach. The proposed approach relies on a Fisher-score-based feature selection method along with a genetic-based extreme gradient boosting (GXGBoost) model in order to determine the most relevant features and to detect IoT botnet attacks. The Fisher score is a representative filter-based feature selection method used to determine significant features and discard irrelevant features through the minimization of intra-class distance and the maximization of inter-class distance. On the other hand, GXGBoost is an optimal and effective model, used to classify the IoT botnet attacks. Several experiments were conducted on a public botnet dataset of IoT devices. The evaluation results obtained using holdout and 10-fold cross-validation techniques showed that the proposed approach had a high detection rate using only three out of the 115 data traffic features and improved the overall performance of the IoT botnet attack detection process. [ABSTRACT FROM AUTHOR]

Although fingerprint-based systems are the commonly used biometric systems, they suffer from a critical vulnerability to a presentation attack (PA). Therefore, several approaches based on a fingerprint biometrics have been developed to increase the robustness against a PA. We propose an alternative approach based on the combination of fingerprint and electrocardiogram (ECG) signals. An ECG signal has advantageous characteristics that prevent the replication. Combining a fingerprint with an ECG signal is a potentially interesting solution to reduce the impact of PAs in biometric systems. We also propose a novel end-to-end deep learning-based fusion neural architecture between a fingerprint and an ECG signal to improve PA detection in fingerprint biometrics. Our model uses state-of-the-art EfficientNets for generating a fingerprint feature representation. For the ECG, we investigate three different architectures based on fully-connected layers (FC), a 1D-convolutional neural network (1D-CNN), and a 2D-convolutional neural network (2D-CNN). The 2D-CNN converts the ECG signals into an image and uses inverted Mobilenet-v2 layers for feature generation. We evaluated the method on a multimodal dataset, that is, a customized fusion of the LivDet 2015 fingerprint dataset and ECG data from real subjects. Experimental results reveal that this architecture yields a better average classification accuracy compared to a single fingerprint modality. [ABSTRACT FROM AUTHOR]

An Intrusion detection system is an essential security tool for protecting services and infrastructures of wireless sensor networks from unseen and unpredictable attacks. Few works of machine learning have been proposed for intrusion detection in wireless sensor networks and that have achieved reasonable results. However, these works still need to be more accurate and efficient against imbalanced data problems in network traffic. In this paper, we proposed a new model to detect intrusion attacks based on a genetic algorithm and an extreme gradient boosting (XGBoot) classifier, called GXGBoost model. The latter is a gradient boosting model designed for improving the performance of traditional models to detect minority classes of attacks in the highly imbalanced data traffic of wireless sensor networks. A set of experiments were conducted on wireless sensor network-detection system (WSN-DS) dataset using holdout and 10 fold cross validation techniques. The results of 10 fold cross validation tests revealed that the proposed approach outperformed the state-of-the-art approaches and other ensemble learning classifiers with high detection rates of 98.2%, 92.9%, 98.9%, and 99.5% for flooding, scheduling, grayhole, and blackhole attacks, respectively, in addition to 99.9% for normal traffic. [ABSTRACT FROM AUTHOR]

Genomic copy number variations (CNVs) are among the most important structural variations. They are linked to several diseases and cancer types. Cancer is a leading cause of death worldwide. Several studies were conducted to investigate the causes of cancer and its association with genomic changes to enhance its management and improve the treatment opportunities. Classification of cancer types based on the CNVs falls in this category of research. We reviewed the recent, most successful methods that used machine learning algorithms to solve this problem and obtained a dataset that was tested by some of these methods for evaluation and comparison purposes. We propose three deep learning techniques to classify cancer types based on CNVs: a six-layer convolutional net (CNN6), residual six-layer convolutional net (ResCNN6), and transfer learning of pretrained VGG16 net. The results of the experiments performed on the data of six cancer types demonstrated a high accuracy of 86% for ResCNN6 followed by 85% for CNN6 and 77% for VGG16. The results revealed a lower prediction accuracy for one of the classes (uterine corpus endometrial carcinoma (UCEC)). Repeating the experiments after excluding this class reveals improvements in the accuracies: 91% for CNN6 and 92% for Res CNN6. We observed that UCEC and ovarian serous carcinoma (OV) share a considerable subset of their features, which causes a struggle for learning in the classifiers. We repeated the experiment again by balancing the six classes through oversampling of the training dataset and the result was an enhancement in both overall and UCEC classification accuracies. [ABSTRACT FROM AUTHOR]

This paper presents an adaptation of the flying ant colony optimization (FACO) algorithm to solve the traveling salesman problem (TSP). This new modification is called dynamic flying ant colony optimization (DFACO). FACO was originally proposed to solve the quality of service (QoS)-aware web service selection problem. Many researchers have addressed the TSP, but most solutions could not avoid the stagnation problem. In FACO, a flying ant deposits a pheromone by injecting it from a distance; therefore, not only the nodes on the path but also the neighboring nodes receive the pheromone. The amount of pheromone a neighboring node receives is inversely proportional to the distance between it and the node on the path. In this work, we modified the FACO algorithm to make it suitable for TSP in several ways. For example, the number of neighboring nodes that received pheromones varied depending on the quality of the solution compared to the rest of the solutions. This helped to balance the exploration and exploitation strategies. We also embedded the 3-Opt algorithm to improve the solution by mitigating the effect of the stagnation problem. Moreover, the colony contained a combination of regular and flying ants. These modifications aim to help the DFACO algorithm obtain better solutions in less processing time and avoid getting stuck in local minima. This work compared DFACO with (1) ACO and five different methods using 24 TSP datasets and (2) parallel ACO (PACO)-3Opt using 22 TSP datasets. The empirical results showed that DFACO achieved the best results compared with ACO and the five different methods for most of the datasets (23 out of 24) in terms of the quality of the solutions. Further, it achieved better results compared with PACO-3Opt for most of the datasets (20 out of 21) in terms of solution quality and execution time. [ABSTRACT FROM AUTHOR]