Search

Over the last few years, a plethora of papers presenting machine-learning-based approaches for intrusion detection have been published. However, the majority of those papers do not compare their results with a proper baseline of a signature-based intrusion detection system, thus violating good machine learning practices. In order to evaluate the pros and cons of the machine-learning-based approach, we replicated a research study that uses a deep neural network model for intrusion detection. The results of our replicated research study expose several systematic problems with the used datasets and evaluation methods. In our experiments, a signature-based intrusion detection system with a minimal setup was able to outperform the tested model even under small traffic changes. Testing the replicated neural network on a new dataset recorded in the same environment with the same attacks using the same tools showed that the accuracy of the neural network dropped to 54%. Furthermore, the often-claimed advantage of being able to detect zero-day attacks could not be seen in our experiments. [ABSTRACT FROM AUTHOR]

Recent technological developments have enabled computers to identify and categorize facial expressions to determine a person's emotional state in an image or a video. This process, called "Facial Expression Recognition (FER)", has become one of the most popular research areas in computer vision. In recent times, deep FER systems have primarily concentrated on addressing two significant challenges: the problem of overfitting due to limited training data availability, and the presence of expression-unrelated variations, including illumination, head pose, image resolution, and identity bias. In this paper, a comprehensive survey is provided on deep FER, encompassing algorithms and datasets that offer insights into these intrinsic problems. Initially, this paper presents a detailed timeline showcasing the evolution of methods and datasets in deep facial expression recognition (FER). This timeline illustrates the progression and development of the techniques and data resources used in FER. Then, a comprehensive review of FER methods is introduced, including the basic principles of FER (components such as preprocessing, feature extraction and classification, and methods, etc.) from the pro-deep learning era (traditional methods using handcrafted features, i.e., SVM and HOG, etc.) to the deep learning era. Moreover, a brief introduction is provided related to the benchmark datasets (there are two categories: controlled environments (lab) and uncontrolled environments (in the wild)) used to evaluate different FER methods and a comparison of different FER models. Existing deep neural networks and related training strategies designed for FER, based on static images and dynamic image sequences, are discussed. The remaining challenges and corresponding opportunities in FER and the future directions for designing robust deep FER systems are also pinpointed. [ABSTRACT FROM AUTHOR]

In this paper, a new method that computes the aesthetics of a melody fragment is proposed, starting from dissonances. While music generated with artificial intelligence applications may be produced considerably more quickly than human-composed music, it has the drawback of not being appreciated like a human composition, being many times perceived by humans as artificial. For achieving supervised machine learning objectives of improving the quality of the great number of generated melodies, it is a challenge to ask humans to grade them. Therefore, it would be preferable if the aesthetics of artificial-intelligence-generated music is calculated by an algorithm. The proposed method in this paper is based on a neural network and a mathematical formula, which has been developed with the help of a study in which 108 students evaluated the aesthetics of several melodies. For evaluation, numerical values generated by this method were compared with ratings provided by human listeners from a second study in which 30 students participated and scores were generated by an existing different method developed by psychologists and three other methods developed by musicians. Our method achieved a Pearson correlation of 0.49 with human aesthetic scores, which is a much better result than other methods obtained. Additionally, our method made a distinction between human-composed melodies and artificial-intelligence-generated scores in the same way that human listeners did. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel attack on image encryption for privacy-preserving deep neural networks (DNNs). Although several encryption schemes have been proposed for privacy-preserving DNNs, existing cipher-text-only attacks (COAs) have succeeded in restoring visual information from encrypted images. Image encryption using the Vision Transformer (ViT) is known to be robust against existing COAs due to the operations of block scrambling and pixel shuffling, which permute divided blocks and pixels in an encrypted image. However, the correlation between blocks in the encrypted image can still be exploited for reconstruction. Therefore, in this paper, a novel jigsaw puzzle solver-based attack that utilizes block correlation is proposed to restore visual information from encrypted images. In the experiments, we evaluated the security of image encryption for privacy-preserving deep neural networks using both conventional and proposed COAs. The experimental results demonstrate that the proposed attack is able to restore almost all visual information from images encrypted for being applied to ViTs. [ABSTRACT FROM AUTHOR]

Vision-based perception modules are increasingly deployed in many applications, especially autonomous vehicles and intelligent robots. These modules are being used to acquire information about the surroundings and identify obstacles. Hence, accurate detection and classification are essential to reach appropriate decisions and take appropriate and safe actions at all times. Current studies have demonstrated that "printed adversarial attacks", known as physical adversarial attacks, can successfully mislead perception models such as object detectors and image classifiers. However, most of these physical attacks are based on noticeable and eye-catching patterns for generated perturbations making them identifiable/detectable by the human eye, in-field tests, or in test drives. In this paper, we propose a camera-based inconspicuous adversarial attack (AdvRain) capable of fooling camera-based perception systems over all objects of the same class. Unlike mask-based FakeWeather attacks that require access to the underlying computing hardware or image memory, our attack is based on emulating the effects of a natural weather condition (i.e., Raindrops) that can be printed on a translucent sticker, which is externally placed over the lens of a camera whenever an adversary plans to trigger an attack. Note, such perturbations are still inconspicuous in real-world deployments and their presence goes unnoticed due to their association with a natural phenomenon. To accomplish this, we develop an iterative process based on performing a random search aiming to identify critical positions to make sure that the performed transformation is adversarial for a target classifier. Our transformation is based on blurring predefined parts of the captured image corresponding to the areas covered by the raindrop. We achieve a drop in average model accuracy of more than 45 % and 40 % on VGG19 for ImageNet dataset and Resnet34 for Caltech-101 dataset, respectively, using only 20 raindrops. [ABSTRACT FROM AUTHOR]

Stroke remains a predominant cause of mortality and disability worldwide. The endeavor to diagnose stroke through biomechanical time-series data coupled with Artificial Intelligence (AI) poses a formidable challenge, especially amidst constrained participant numbers. The challenge escalates when dealing with small datasets, a common scenario in preliminary medical research. While recent advances have ushered in few-shot learning algorithms adept at handling sparse data, this paper pioneers a distinctive methodology involving a visualization-centric approach to navigating the small-data challenge in diagnosing stroke survivors based on gait-analysis-derived biomechanical data. Employing Siamese neural networks (SNNs), our method transforms a biomechanical time series into visually intuitive images, facilitating a unique analytical lens. The kinematic data encapsulated comprise a spectrum of gait metrics, including movements of the ankle, knee, hip, and center of mass in three dimensions for both paretic and non-paretic legs. Following the visual transformation, the SNN serves as a potent feature extractor, mapping the data into a high-dimensional feature space conducive to classification. The extracted features are subsequently fed into various machine learning (ML) models like support vector machines (SVMs), Random Forest (RF), or neural networks (NN) for classification. In pursuit of heightened interpretability, a cornerstone in medical AI applications, we employ the Grad-CAM (Class Activation Map) tool to visually highlight the critical regions influencing the model's decision. Our methodology, though exploratory, showcases a promising avenue for leveraging visualized biomechanical data in stroke diagnosis, achieving a perfect classification rate in our preliminary dataset. The visual inspection of generated images elucidates a clear separation of classes (100%), underscoring the potential of this visualization-driven approach in the realm of small data. This proof-of-concept study accentuates the novelty of visual data transformation in enhancing both interpretability and performance in stroke diagnosis using limited data, laying a robust foundation for future research in larger-scale evaluations. [ABSTRACT FROM AUTHOR]

Few-shot learning is a deep learning subfield that is the focus of research nowadays. This paper addresses the research question of whether a triplet-trained Siamese network, initially designed for multi-class classification, can effectively handle multi-label classification. We conducted a case study to identify any limitations in its application. The experiments were conducted on a dataset containing Hungarian legal decisions of administrative agencies in tax matters belonging to a major legal content provider. We also tested how different Siamese embeddings compare on classifying a previously non-existing label on a binary and a multi-label setting. We found that triplet-trained Siamese networks can be applied to perform classification but with a sampling restriction during training. We also found that the overlap between labels affects the results negatively. The few-shot model, seeing only ten examples for each label, provided competitive results compared to models trained on tens of thousands of court decisions using tf-idf vectorization and logistic regression. [ABSTRACT FROM AUTHOR]

Radial pulse signals are produced by the periodic ejection of blood from the heart, and physiological and pathological information of the human body can be analyzed by extracting the time-domain characteristics of pulse waves. However, since pulse signals are weak physiological signals on the body surface and complex, the acquisition of pulse characteristics using the traditional curvature method will produce a large error, which cannot meet the needs of pulse wave analysis in current clinical practice. To solve this problem, a multi-morphological pulse signal feature recognition algorithm based on the one-dimensional deep convolutional neural network (1D-DCNN) model is proposed. We used the multi-channel pulse diagnosis instrument independently developed by the team to collect radial pulse signals under continuous pressure of the test subjects and collected 115 subjects and extracted a total of 1300 single-cycle pulse signals and then divided these pulse signals into 6 different forms. Five types of pulse signal time-domain feature points were labeled, and five independent feature point datasets were labeled and formed five customized neural network models that were generated to train and identify the pulse feature point datasets independently. The results show that the correction coefficient ( R a d j u s t e d 2 ) of the multi-class pulse signal processing algorithm proposed in this paper for each type of feature point recognition reaches more than 0.92. The performance is significantly better than that of the traditional curvature method, which shows the accuracy and superiority of the proposed method. Therefore, the multi-class pulse signal characteristic parameter recognition model based on the 1D-DCNN model proposed in this paper can efficiently and accurately identify pulse time-domain characteristic parameters, which can be applied to discriminate time-domain pulse information in clinical practice and assist doctors in diagnosis. [ABSTRACT FROM AUTHOR]

In recent years, rapid advancements in both hardware and software technologies have resulted in the ability to execute artificial intelligence (AI) algorithms on low-resource devices. The combination of high-speed, low-power electronic hardware and efficient AI algorithms is driving the emergence of on-device AI. Deep neural networks (DNNs) are highly effective AI algorithms used for identifying patterns in complex data. DNNs, however, contain many parameters and operations that make them computationally intensive to execute. Accordingly, DNNs are usually executed on high-resource backend processors. This causes an increase in data processing latency and energy expenditure. Therefore, modern strategies are being developed to facilitate the implementation of DNNs on devices with limited resources. This paper presents a detailed review of the current methods and structures that have been developed to deploy DNNs on devices with limited resources. Firstly, an overview of DNNs is presented. Next, the methods used to implement DNNs on resource-constrained devices are explained. Following this, the existing works reported in the literature on the execution of DNNs on low-resource devices are reviewed. The reviewed works are classified into three categories: software, hardware, and hardware/software co-design. Then, a discussion on the reviewed approaches is given, followed by a list of challenges and future prospects of on-device AI, together with its emerging applications. [ABSTRACT FROM AUTHOR]

The paper presents the data on lipid fraction extraction from the raw fat of hibernating hunting animals. The processing of valuable raw materials must be maximized. For this purpose, various methods of rendering are used. As a result of temperature exposure, the protein part of raw fat undergoes significant changes. The protein denatures under the influence of temperature, and the dross formed during the rendering process absorbs and retains up to 30% of the fat. The authors propose using proteolytic enzyme preparations for a more complete extraction of fats, as the enzymes will hydrolyze the protein into compounds of lower molecular weight both before and during the rendering process. The experiment proved that the biocatalytic method allows achieving a fat yield of more than 95%. The best result can be obtained if the rendering is carried out at optimal parameters, which can be defined using a mathematical model. Mathematical modeling was carried out using an artificial neural network. During the study, a fully connected neural network was designed; it had eight hidden layers with 64 neurons in each, and its accuracy was measured by mean relative error, which amounted to 5.16%. With the help of the network, the optimal values of applied concentration, temperature and duration of rendering, at which a fat yield of more than 98% is achieved, were determined for each enzyme preparation. After that, the obtained values were confirmed experimentally. Thus, the study showed the efficiency of using artificial neural networks for modeling the biocatalytic method of lipid extraction. [ABSTRACT FROM AUTHOR]

Speech separation is a well-known problem, especially when there is only one sound mixture available. Estimating the Ideal Binary Mask (IBM) is one solution to this problem. Recent research has focused on the supervised classification approach. The challenge of extracting features from the sources is critical for this method. Speech separation has been accomplished by using a variety of feature extraction models. The majority of them, however, are concentrated on a single feature. The complementary nature of various features have not been thoroughly investigated. In this paper, we propose a deep neural network (DNN) ensemble architecture to completely explore the complimentary nature of the diverse features obtained from raw acoustic features. We examined the penultimate discriminative representations instead of employing the features acquired from the output layer. The learned representations were also fused to produce a new features vector, which was then classified by using the Extreme Learning Machine (ELM). In addition, a genetic algorithm (GA) was created to optimize the parameters globally. The results of the experiments showed that our proposed system completely considered various features and produced a high-quality IBM under different conditions. [ABSTRACT FROM AUTHOR]

Deep neural networks have proven to be able to learn rich internal representations, including for features that can also be used for different purposes than those the networks are originally developed for. In this paper, we are interested in exploring such ability and, to this aim, we propose a novel approach for investigating the internal behavior of networks trained for source code processing tasks. Using a simple autoencoder trained in the reconstruction of vectors representing programs (i.e., program embeddings), we first analyze the performance of the internal neurons in classifying programs according to different labeling policies inspired by real programming issues, showing that some neurons can actually detect different program properties. We then study the dynamics of the network from an information-theoretic standpoint, namely by considering the neurons as signaling systems and by computing the corresponding entropy. Further, we define a way to distinguish neurons according to their behavior, to consider them as formally associated with different abstract concepts, and through the application of nonparametric statistical tests to pairs of neurons, we look for neurons with unique (or almost unique) associated concepts, showing that the entropy value of a neuron is related to the rareness of its concept. Finally, we discuss how the proposed approaches for ranking the neurons can be generalized to different domains and applied to more sophisticated and specialized networks so as to help the research in the growing field of explainable artificial intelligence. [ABSTRACT FROM AUTHOR]

In recent years, social platforms have become integrated in a variety of economic, political and cultural domains. Social media have become the primary outlets for many citizens to consume news and information, and, at the same time, to produce and share online a large amount of data and meta-data. This paper presents an innovative system able to analyze visual information shared by citizens on social media during extreme events for contributing to the situational awareness and supporting people in charge of coordinating the emergency management. The system analyzes all posts containing images shared by users by taking into account: (a) the event class and (b) the GPS coordinates of the geographical area affected by the event. Then, a Single Shot Multibox Detector (SSD) network is applied to select only the posted images correctly related to the event class and an advanced image processing procedure is used to verify if these images are correlated with the geographical area where the emergency event is ongoing. Several experiments have been carried out to evaluate the performance of the proposed system in the context of different emergency situations caused by earthquakes, floods and terrorist attacks. [ABSTRACT FROM AUTHOR]

Drowning is one of the top five causes of death for children aged 1–14 worldwide. According to data from the World Health Organization (WHO), drowning is the third most common reason for unintentional fatalities. Designing a drowning detection system is becoming increasingly necessary in order to ensure the safety of swimmers, particularly children. This paper presents a computer vision and deep learning-based early drowning detection approach. We utilized five convolutional neural network models and trained them on our data. These models are SqueezeNet, GoogleNet, AlexNet, ShuffleNet, and ResNet50. ResNet50 showed the best performance, as it achieved 100% prediction accuracy with a reasonable training time. When compared to other approaches, the proposed approach performed exceptionally well in terms of prediction accuracy and computational cost. [ABSTRACT FROM AUTHOR]

Recently, approaches based on the transformation of tabular data into images have gained a lot of scientific attention. This is explained by the fact that convolutional neural networks (CNNs) have shown good results in computer vision and other image-based classification tasks. Transformation of features without spatial relations to images allows the application of deep neural networks to a wide range of analysis tasks. This paper analyzes existing approaches to feature transformation based on the conversion of the features of network traffic into images and discusses their advantages and disadvantages. The authors also propose an approach to the transformation of raw network packets into images and analyze its efficiency in the task of network attack detection in a cyber-physical object, including its robustness to novel and unseen attacks. [ABSTRACT FROM AUTHOR]

When analyzing a financial asset, it is essential to study the trend of its time series. It is also necessary to examine its evolution and activity over time to statistically analyze its possible future behavior. Both retail and institutional investors base their trading strategies on these analyses. One of the most used techniques to study financial time series is to analyze its dynamic structure using auto-regressive models, simple moving average models (SMA), and mixed auto-regressive moving average models (ARMA). These techniques, unfortunately, do not always provide appreciable results both at a statistical level and as the Risk-Reward Ratio (RRR); above all, each system has its pros and cons. In this paper, we present CryptoNet; this system is based on the time series extraction exploiting the vast potential of artificial intelligence (AI) and machine learning (ML). Specifically, we focused on time series trends extraction by developing an artificial neural network, trained and tested on two famous crypto-currencies: Bitcoinand Ether. CryptoNet learning algorithm improved the classic linear regression model up to 31% of MAE (mean absolute error). Results from this work should encourage machine learning techniques in sectors classically reluctant to adopt non-standard approaches. [ABSTRACT FROM AUTHOR]

Map matching is a crucial data processing task for transferring measurements from the dynamic sensor location to the relevant road segment. It is especially important when estimating road network speed by using probe vehicles (floating car data) as speed measurement sensors. Most common approaches rely on finding the closet road segment, but road network geometry (e.g., dense areas, two-way streets, and superposition of road segments due to different heights) and inaccuracy in the GNSS location (up to decades of meters in urban areas) can wrongly allocate up to 30% of the measurements. More advanced methods rely on taking the topology of the network into account, significantly improving the accuracy at a higher computational cost, especially when the accuracy of the GNSS location is low. In order to both improve the accuracy of the "closet road segment" methods and reduce the processing time of the topology-based methods, the data can be pre-processed using AI techniques to reduce noise created by the inaccuracy of the GNSS location and improve the overall accuracy of the map-matching task. This paper applies AI to correct GNSS locations and improve the map-matching results, achieving a matching accuracy of 76%. The proposed methodology is demonstrated to the floating car data generated by a fleet of 1200 taxi vehicles in Thessaloniki used to estimate road network speed in real time for information services and for supporting traffic management in the city. [ABSTRACT FROM AUTHOR]

The analysis of influential machine parameters can be useful to plan and design a plastic injection molding process. However, current research in parameter analysis is mostly based on computer-aided engineering (CAE) or simulation which have been demonstrated to be inadequate for analyzing complex behavioral changes in the real injection molding process. More advanced approaches using machine learning technology specifically with artificial neural networks (ANNs) brought promising results in terms of prediction accuracy. Nevertheless, the black box and distributed representation of ANN prevent humans from gaining an insight into which process parameters give a significant influence on the final prediction output. Therefore, in this paper, we develop a simpler ANN model by using structural learning with forgetting (SLF) as the algorithm for the training process. Instead of typical backpropagation which generated a fully connected layer of the ANN model, SLF only reveals the important neurons and connections. Hence, the training process of SLF leaves only influential connections and neurons. Since each of the neurons specifically on the input layer represent each of the injection molding parameters, the ANN-SLF model can be further investigated to determine the influential process parameters. By applying SLF to the ANN training process, this experiment has successfully extracted a set of significant injection molding process parameters. [ABSTRACT FROM AUTHOR]

Speech signals carry various bits of information relevant to the speaker such as age, gender, accent, language, health, and emotions. Emotions are conveyed through modulations of facial and vocal expressions. This paper conducts an empirical comparison of performances between the classical classifiers: Gaussian Mixture Model (GMM), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Artificial neural networks (ANN); and the deep learning classifiers, i.e., Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Gated Recurrent Unit (GRU) in addition to the ivector approach for a text-independent speaker verification task in neutral and emotional talking environments. The deep models undergo hyperparameter tuning using the Grid Search optimization algorithm. The models are trained and tested using a private Arabic Emirati Speech Database, Ryerson Audio–Visual Database of Emotional Speech and Song dataset (RAVDESS) database, and a public Crowd-Sourced Emotional Multimodal Actors (CREMA) database. Experimental results illustrate that deep architectures do not necessarily outperform classical classifiers. In fact, evaluation was carried out through Equal Error Rate (EER) along with Area Under the Curve (AUC) scores. The findings reveal that the GMM model yields the lowest EER values and the best AUC scores across all datasets, amongst classical classifiers. In addition, the ivector model surpasses all the fine-tuned deep models (CNN, LSTM, and GRU) based on both evaluation metrics in the neutral, as well as the emotional speech. In addition, the GMM outperforms the ivector using the Emirati and RAVDESS databases. [ABSTRACT FROM AUTHOR]

The recent boom of artificial Neural Networks (NN) has shown that NN can provide viable solutions to a variety of problems. However, their complexity and the lack of efficient interpretation of NN architectures (commonly considered black box techniques) has adverse effects on the optimization of each NN architecture. One cannot simply use a generic topology and have the best performance in every application field, since the network topology is commonly fine-tuned to the problem/dataset in question. In this paper, we introduce a novel method of computationally assessing the complexity of the dataset. The NN is treated as an information channel, and thus information theory is used to estimate the optimal number of neurons for each layer, reducing the memory and computational load, while achieving the same, if not greater, accuracy. Experiments using common datasets confirm the theoretical findings, and the derived algorithm seems to improve the performance of the original architecture. [ABSTRACT FROM AUTHOR]

Online purchasing has developed rapidly in recent years due to its efficiency, convenience, low cost, and product variety. This has increased the number of online multi-category e-commerce retailers that sell a variety of product categories. Due to the growth in the number of players, each company needs to optimize its own business strategy in order to compete. Customer lifetime value (CLV) is a common metric that multi-category e-commerce retailers usually consider for competition because it helps determine the most valuable customers for the retailers. However, in this paper, we introduce two additional novel factors in addition to CLV to determine which customers will bring in the highest revenue in the future: distinct product category (DPC) and trend in amount spent (TAS). Then, we propose a new framework. We utilized, for the first time in the relevant literature, a multi-output deep neural network (DNN) model to test our proposed framework while forecasting CLV, DPC, and TAS together. To make this outcome applicable in real life, we constructed customer clusters that allow the management of multi-category e-commerce companies to segment end-users based on the three variables. We compared the proposed framework (constructed with multiple outputs: CLV, DPC, and TAS) against a baseline single-output model to determine the combined effect of the multi-output model. In addition, we also compared the proposed model with multi-output Decision Tree (DT) and multi-output Random Forest (RF) algorithms on the same dataset. The results indicate that the multi-output DNN model outperforms the single-output DNN model, multi-output DT, and multi-output RF across all assessment measures, proving that the multi-output DNN model is more suitable for multi-category e-commerce retailers' usage. Furthermore, Shapley values derived through the explainable artificial intelligence method are used to interpret the decisions of the DNN. This practice demonstrates which inputs contribute more to the outcomes (a significant novelty in interpreting the DNN model for the CLV). [ABSTRACT FROM AUTHOR]

Recently, Deep Neural Networks (DNNs) have become a central subject of discussion in computer vision for a broad range of applications, including image classification and face recognition. Compared to existing conventional machine learning methods, deep learning algorithms have shown prominent performance with high accuracy and speed. However, they always require a large amount of data to achieve adequate robustness. Furthermore, additional samples are time-consuming and expensive to collect. In this paper, we propose an approach that combines generative methods and basic manipulations for image data augmentations and the FaceNet model with Support Vector Machine (SVM) for face recognition. To do so, the images were first preprocessed by a Deep Convolutional Generative Adversarial Net (DCGAN) to generate samples having realistic properties inseparable from those of the original datasets. Second, basic manipulations were applied on the images produced by DCGAN in order to increase the amount of training data. Finally, FaceNet was employed as a face recognition model. FaceNet detects faces using MTCNN, 128-D face embedding is computed to quantify each face, and an SVM was used on top of the embeddings for classification. Experiments carried out on the LFW and VGG image databases and ChokePoint video database demonstrate that the combination of basic and generative methods for augmentation boosted face recognition performance, leading to better recognition results. [ABSTRACT FROM AUTHOR]

Productivity losses caused by absenteeism at work cost U.S. employers billions of dollars each year. In addition, employers typically spend a considerable amount of time managing employees who perform poorly. By using predictive analytics and machine learning algorithms, organizations can make better decisions, thereby increasing organizational productivity, reducing costs, and improving efficiency. Thus, in this paper we propose hybrid optimization methods in order to find the most parsimonious model for absenteeism classification. We utilized data from a Brazilian courier company. In order to categorize absenteeism classes, we preprocessed the data, selected the attributes via multiple methods, balanced the dataset using the synthetic minority over-sampling method, and then employed four methods of machine learning classification: Support Vector Machine (SVM), Multinomial Logistic Regression (MLR), Artificial Neural Network (ANN), and Random Forest (RF). We selected the best model based on several validation scores, and compared its performance against the existing model. Furthermore, project managers may lack experience in machine learning, or may not have the time to spend developing machine learning algorithms. Thus, we propose a web-based interactive tool supported by cognitive analytics management (CAM) theory. The web-based decision tool enables managers to make more informed decisions, and can be used without any prior knowledge of machine learning. Understanding absenteeism patterns can assist managers in revising policies or creating new arrangements to reduce absences in the workplace, financial losses, and the probability of economic insolvency. [ABSTRACT FROM AUTHOR]

Field Programmable Gate Array (FPGA) accelerators have been widely adopted for artificial intelligence (AI) applications on edge devices (Edge-AI) utilizing Deep Neural Networks (DNN) architectures. FPGAs have gained their reputation due to the greater energy efficiency and high parallelism than microcontrollers (MCU) and graphical processing units (GPU), while they are easier to develop and more reconfigurable than the Application Specific Integrated Circuit (ASIC). The development and building of AI applications on resource constraint devices such as FPGAs remains a challenge, however, due to the co-design approach, which requires a valuable expertise in low-level hardware design and in software development. This paper explores the efficacy and the dynamic deployment of hardware accelerated applications on the Kria KV260 development platform based on the Xilinx Kria K26 system-on-module (SoM), which includes a Zynq multiprocessor system-on-chip (MPSoC). The platform supports the Python-based PYNQ framework and maintains a high level of versatility with the support of custom bitstreams (overlays). The demonstration proved the reconfigurabibilty and the overall ease of implementation with low-footprint machine learning (ML) algorithms. [ABSTRACT FROM AUTHOR]

There are more than 962 million people aged 60 and up globally. Physical activity declines as people get older, as does their capacity to undertake everyday tasks, effecting both physical and mental health. Many researchers use machine learning and deep learning methods to recognize human activities, but very few studies have been focused on human activity recognition of elderly people. This paper focuses on providing assistance to elderly people by monitoring their activities in different indoor and outdoor environments using gyroscope and accelerometer data collected from a smart phone. Smart phones have been routinely used to monitor the activities of persons with impairments; routine activities such as sitting, walking, going upstairs, going downstairs, standing, and lying are included in the dataset. Conventional Machine Learning and Deep Learning algorithms such as k-Nearest Neighbors, Random Forest, Support Vector Machine, Artificial Neural Network, and Long Short-Term Memory Network are used for human activity recognition. Long Short-Term Memory is a recurrent neural network variation that is best suited to handling temporal sequences. Two-fold and ten-fold cross-validation methods were performed to show the effect of changing the data in the training and testing dataset. Among all the classification techniques, the proposed Long Short-Term Memory Network gave the best accuracy of 95.04%. However, Support Vector Machine gave 89.07% accuracy with a very low computational time of 0.42 min using 10-fold cross-validation. [ABSTRACT FROM AUTHOR]

As an informative electroencephalogram (EEG) signal, steady-state visual evoked potential (SSVEP) stands out from many paradigms for application in wireless wearable devices. However, its data are usually enormous, occupy too many bandwidth sources and require immense power when transmitted in the raw data form, so it is necessary to compress the signal. This paper proposes a personalized EEG compression and reconstruction algorithm for the SSVEP application. In the algorithm, to realize personalization, a primary artificial neural network (ANN) model is first pre-trained with the open benchmark database towards BCI application (BETA). Then, an adaptive ANN model is generated with incremental learning for each subject to compress their individual data. Additionally, a personalized, non-uniform quantization method is proposed to reduce the errors caused by compression. The recognition accuracy only decreases by 3.79% when the compression rate is 12.7 times, and is tested on BETA. The proposed algorithm can reduce signal loss by from 50.43% to 81.08% in the accuracy test compared to the case without ANN and uniform quantization. [ABSTRACT FROM AUTHOR]

Identification of the malignancy of tissues from Histopathological images has always been an issue of concern to doctors and radiologists. This task is time-consuming, tedious and moreover very challenging. Success in finding malignancy from Histopathological images primarily depends on long-term experience, though sometimes experts disagree on their decisions. However, Computer Aided Diagnosis (CAD) techniques help the radiologist to give a second opinion that can increase the reliability of the radiologist's decision. Among the different image analysis techniques, classification of the images has always been a challenging task. Due to the intense complexity of biomedical images, it is always very challenging to provide a reliable decision about an image. The state-of-the-art Convolutional Neural Network (CNN) technique has had great success in natural image classification. Utilizing advanced engineering techniques along with the CNN, in this paper, we have classified a set of Histopathological Breast-Cancer (BC) images utilizing a state-of-the-art CNN model containing a residual block. Conventional CNN operation takes raw images as input and extracts the global features; however, the object oriented local features also contain significant information--for example, the Local Binary Pattern (LBP) represents the effective textural information, Histogram represent the pixel strength distribution, Contourlet Transform (CT) gives much detailed information about the smoothness about the edges, and Discrete Fourier Transform (DFT) derives frequency-domain information from the image. Utilizing these advantages, along with our proposed novel CNN model, we have examined the performance of the novel CNN model as Histopathological image classifier. To do so, we have introduced five cases: (a) Convolutional Neural Network Raw Image (CNN-I); (b) Convolutional Neural Network CT Histogram (CNN-CH); (c) Convolutional Neural Network CT LBP (CNN-CL); (d) Convolutional Neural Network Discrete Fourier Transform (CNN-DF); (e) Convolutional Neural Network Discrete Cosine Transform (CNN-DC). We have performed our experiments on the BreakHis image dataset. The best performance is achieved when we utilize the CNN-CH model on a 200× dataset that provides Accuracy, Sensitivity, False Positive Rate, False Negative Rate, Recall Value, Precision and F-measure of 92.19%, 94.94%, 5.07%, 1.70%, 98.20%, 98.00% and 98.00%, respectively. [ABSTRACT FROM AUTHOR]

The traditional algorithms for recognizing handwritten alphanumeric characters are dependent on hand-designed features. In recent days, deep learning techniques have brought about new breakthrough technology for pattern recognition applications, especially for handwritten recognition. However, deeper networks are needed to deliver state-of-the-art results in this area. In this paper, inspired by the success of the very deep state-of-the-art VGGNet, we propose Alphanumeric VGG net for Arabic handwritten alphanumeric character recognition. Alphanumeric VGG net is constructed by thirteen convolutional layers, two max-pooling layers, and three fully-connected layers. The proposed model is fast and reliable, which improves the classification performance. Besides, this model has also reduced the overall complexity of VGGNet. We evaluated our approach on two benchmarking databases. We have achieved very promising results, with a validation accuracy of 99.66% for the ADBase database and 97.32% for the HACDB database. [ABSTRACT FROM AUTHOR]

Evolutionary information theory is a constructive approach that studies information in the context of evolutionary processes, which are ubiquitous in nature and society. In this paper, we develop foundations of evolutionary information theory, building several measures of evolutionary information and obtaining their properties. These measures are based on mathematical models of evolutionary computations, machines and automata. To measure evolutionary information in an invariant form, we construct and study universal evolutionary machines and automata, which form the base for evolutionary information theory. The first class of measures introduced and studied in this paper is evolutionary information size of symbolic objects relative to classes of automata or machines. In particular, it is proved that there is an invariant and optimal evolutionary information size relative to different classes of evolutionary machines. As a rule, different classes of algorithms or automata determine different information size for the same object. The more powerful classes of algorithms or automata decrease the information size of an object in comparison with the information size of an object relative to weaker4 classes of algorithms or machines. The second class of measures for evolutionary information in symbolic objects is studied by introduction of the quantity of evolutionary information about symbolic objects relative to a class of automata or machines. To give an example of applications, we briefly describe a possibility of modeling physical evolution with evolutionary machines to demonstrate applicability of evolutionary information theory to all material processes. At the end of the paper, directions for future research are suggested. [ABSTRACT FROM AUTHOR]

Near-infrared (NIR) face recognition has attracted increasing attention because of its advantage of illumination invariance. However, traditional face recognition methods based on NIR are designed for and tested in cooperative-user applications. In this paper, we present a convolutional neural network (CNN) for NIR face recognition (specifically face identification) in non-cooperative-user applications. The proposed NIRFaceNet is modified from GoogLeNet, but has a more compact structure designed specifically for the Chinese Academy of Sciences Institute of Automation (CASIA) NIR database and can achieve higher identification rates with less training time and less processing time. The experimental results demonstrate that NIRFaceNet has an overall advantage compared to other methods in the NIR face recognition domain when image blur and noise are present. The performance suggests that the proposed NIRFaceNet method may be more suitable for non-cooperative-user applications. [ABSTRACT FROM AUTHOR]

In this paper, we incorporate a GARCH model into an artificial neural network (ANN) for financial volatility modeling and estimate the parameters in Tensorflow. Our goal was to better predict stock volatility. We evaluate the performance of the models using the mean absolute errors of powers of the out-of-sample returns between 2 March 2018 and 28 February 2020. Our results show that our modeling procedure with an ANN can outperform the standard GARCH(1,1) model with standardized Student's t distribution. Our variable importance analysis shows that Net Debt/EBITA is among the six most important predictor variables in all of the neural network models we have examined. The main contribution of this paper is that we propose a Long Short-Term Memory (LSTM) model with a GARCH framework because LSTM can systematically take into consideration potential nonlinearity in volatility structure at different time points. One of the advantages of our research is that the proposed models are easy to implement because our proposed models can be run in Tensorflow, a Python package that enables fast and automatic optimization. Another advantage is that the proposed models enable variable importance analysis. [ABSTRACT FROM AUTHOR]

Traditional time series forecasting techniques can not extract good enough sequence data features, and their accuracies are limited. The deep learning structure SeriesNet is an advanced method, which adopts hybrid neural networks, including dilated causal convolutional neural network (DC-CNN) and Long-short term memory recurrent neural network (LSTM-RNN), to learn multi-range and multi-level features from multi-conditional time series with higher accuracy. However, they didn't consider the attention mechanisms to learn temporal features. Besides, the conditioning method for CNN and RNN is not specific, and the number of parameters in each layer is tremendous. This paper proposes the conditioning method for two types of neural networks, and respectively uses the gated recurrent unit network (GRU) and the dilated depthwise separable temporal convolutional networks (DDSTCNs) instead of LSTM and DC-CNN for reducing the parameters. Furthermore, this paper presents the lightweight RNN-based hidden state attention module (HSAM) combined with the proposed CNN-based convolutional block attention module (CBAM) for time series forecasting. Experimental results show our model is superior to other models from the viewpoint of forecasting accuracy and computation efficiency. [ABSTRACT FROM AUTHOR]

Most of the existing research papers study the emotion recognition of Minnan songs from the perspectives of music analysis theory and music appreciation. However, these investigations do not explore any possibility of carrying out an automatic emotion recognition of Minnan songs. In this paper, we propose a model that consists of four main modules to classify the emotion of Minnan songs by using the bimodal data—song lyrics and audio. In the proposed model, an attention-based Long Short-Term Memory (LSTM) neural network is applied to extract lyrical features, and a Convolutional Neural Network (CNN) is used to extract the audio features from the spectrum. Then, two kinds of extracted features are concatenated by multimodal compact bilinear pooling, and finally, the concatenated features are input to the classifying module to determine the song emotion. We designed three experiment groups to investigate the classifying performance of combinations of the four main parts, the comparisons of proposed model with the current approaches and the influence of a few key parameters on the performance of emotion recognition. The results show that the proposed model exhibits better performance over all other experimental groups. The accuracy, precision and recall of the proposed model exceed 0.80 in a combination of appropriate parameters. [ABSTRACT FROM AUTHOR]

In this paper, the viability of neural network implementations of core technologies (the focus of this paper is on text technologies) for 10 resource-scarce South African languages is evaluated. Neural networks are increasingly being used in place of other machine learning methods for many natural language processing tasks with good results. However, in the South African context, where most languages are resource-scarce, very little research has been done on neural network implementations of core language technologies. In this paper, we address this gap by evaluating neural network implementations of four core technologies for ten South African languages. The technologies we address are part of speech tagging, named entity recognition, compound analysis and lemmatization. Neural architectures that performed well on similar tasks in other settings were implemented for each task and the performance was assessed in comparison with currently used machine learning implementations of each technology. The neural network models evaluated perform better than the baselines for compound analysis, are viable and comparable to the baseline on most languages for POS tagging and NER, and are viable, but not on par with the baseline, for Afrikaans lemmatization. [ABSTRACT FROM AUTHOR]

As an important part of emotion research, facial expression recognition is a necessary requirement in human–machine interface. Generally, a face expression recognition system includes face detection, feature extraction, and feature classification. Although great success has been made by the traditional machine learning methods, most of them have complex computational problems and lack the ability to extract comprehensive and abstract features. Deep learning-based methods can realize a higher recognition rate for facial expressions, but a large number of training samples and tuning parameters are needed, and the hardware requirement is very high. For the above problems, this paper proposes a method combining features that extracted by the convolutional neural network (CNN) with the C4.5 classifier to recognize facial expressions, which not only can address the incompleteness of handcrafted features but also can avoid the high hardware configuration in the deep learning model. Considering some problems of overfitting and weak generalization ability of the single classifier, random forest is applied in this paper. Meanwhile, this paper makes some improvements for C4.5 classifier and the traditional random forest in the process of experiments. A large number of experiments have proved the effectiveness and feasibility of the proposed method. [ABSTRACT FROM AUTHOR]

Automatic Classification of Wireless Signals (ACWS), which is an intermediate step between signal detection and demodulation, is investigated in this paper. ACWS plays a crucial role in several military and non-military applications, by identifying interference sources and adversary attacks, to achieve efficient radio spectrum management. The performance of traditional feature-based (FB) classification approaches is limited due to their specific input feature set, which in turn results in poor generalization under unknown conditions. Therefore, in this paper, a novel feature-based classifier Neural-Induced Support Vector Machine (NSVM) is proposed, in which the features are learned automatically from raw input signals using Convolutional Neural Networks (CNN). The output of NSVM is given by a Gaussian Support Vector Machine (SVM), which takes the features learned by CNN as its input. The proposed scheme NSVM is trained as a single architecture, and in this way, it learns to minimize a margin-based loss instead of cross-entropy loss. The proposed scheme NSVM outperforms the traditional softmax-based CNN modulation classifier by managing faster convergence of accuracy and loss curves during training. Furthermore, the robustness of the NSVM classifier is verified by extensive simulation experiments under the presence of several non-ideal real-world channel impairments over a range of signal-to-noise ratio (SNR) values. The performance of NSVM is remarkable in classifying wireless signals, such as at low signal-to-noise ratio (SNR), the overall averaged classification accuracy is > 97% at SNR = −2 dB and at higher SNR it achieves overall classification accuracy at > 99%, when SNR = 10 dB. In addition to that, the analytical comparison with other studies shows the performance of NSVM is superior over a range of settings. [ABSTRACT FROM AUTHOR]

Individual identification of dairy cows based on computer vision technology shows strong performance and practicality. Accurate identification of each dairy cow is the prerequisite of artificial intelligence technology applied in smart animal husbandry. While the rump of each dairy cow also has lots of important features, so do the back and head, which are also important for individual recognition. In this paper, we propose a non-contact cow rump identification method based on convolutional neural networks. First, the rump image sequences of the cows while feeding were collected. Then, an object detection model was applied to detect the cow rump object in each frame of image. Finally, a fine-tuned convolutional neural network model was trained to identify cow rumps. An image dataset containing 195 different cows was created to validate the proposed method. The method achieved an identification accuracy of 99.76%, which showed a better performance compared to other related methods and a good potential in the actual production environment of cow husbandry, and the model is light enough to be deployed in an edge-computing device. [ABSTRACT FROM AUTHOR]

Yield prediction for tomatoes in greenhouses is an important basis for making production plans, and yield prediction accuracy directly affects economic benefits. To improve the prediction accuracy of tomato yield in Chinese-style solar greenhouses (CSGs), a wavelet neural network (WNN) model optimized by a genetic algorithm (GA-WNN) is applied. Eight variables are selected as input parameters and the tomato yield is the prediction output. The GA is adopted to optimize the initial weights, thresholds, and translation factors of the WNN. The experiment results show that the mean relative errors (MREs) of the GA-WNN model, WNN model, and backpropagation (BP) neural network model are 0.0067, 0.0104, and 0.0242, respectively. The results root mean square errors (RMSEs) are 1.725, 2.520, and 5.548, respectively. The EC values are 0.9960, 0.9935, and 0.9868, respectively. Therefore, the GA-WNN model has a higher prediction precision and a better fitting ability compared with the BP and the WNN prediction models. The research of this paper is useful from both theoretical and technical perspectives for quantitative tomato yield prediction in the CSGs. [ABSTRACT FROM AUTHOR]

Sampling is an important step in the machine learning process because it prioritizes samples that help the model best summarize the important concepts required for the task at hand. The process of determining the best sampling method has been rarely studied in the context of graph neural networks. In this paper, we evaluate multiple sampling methods (i.e., ascending and descending) that sample based off different definitions of centrality (i.e., Voterank, Pagerank, degree) to observe its relation with network topology. We find that no sampling method is superior across all network topologies. Additionally, we find situations where ascending sampling provides better classification scores, showing the strength of weak ties. Two strategies are then created to predict the best sampling method, one that observes the homogeneous connectivity of the nodes, and one that observes the network topology. In both methods, we are able to evaluate the best sampling direction consistently. [ABSTRACT FROM AUTHOR]

Transfer learning is one of the popular methods for solving the problem that the models built on the source domain cannot be directly applied to the target domain in the cross-domain sentiment classification. This paper proposes a transfer learning method based on the multi-layer convolutional neural network (CNN). Interestingly, we construct a convolutional neural network model to extract features from the source domain and share the weights in the convolutional layer and the pooling layer between the source and target domain samples. Next, we fine-tune the weights in the last layer, named the fully connected layer, and transfer the models from the source domain to the target domain. Comparing with the classical transfer learning methods, the method proposed in this paper does not need to retrain the network for the target domain. The experimental evaluation of the cross-domain data set shows that the proposed method achieves a relatively good performance. [ABSTRACT FROM AUTHOR]

Malware creators generate new malicious software samples by making minor changes in previously generated code, in order to reuse malicious code, as well as to go unnoticed from signature-based antivirus software. As a result, various families of variations of the same initial code exist today. Visualization of compiled executables for malware analysis has been proposed several years ago. Visualization can greatly assist malware classification and requires neither disassembly nor code execution. Moreover, new variations of known malware families are instantly detected, in contrast to traditional signature-based antivirus software. This paper addresses the problem of identifying variations of existing malware visualized as images. A new malware detection system based on a two-level Artificial Neural Network (ANN) is proposed. The classification is based on file and image features. The proposed system is tested on the 'Malimg' dataset consisting of the visual representation of well-known malware families. From this set some important image features are extracted. Based on these features, the ANN is trained. Then, this ANN is used to detect and classify other samples of the dataset. Malware families creating a confusion are classified by a second level of ANNs. The proposed two-level ANN method excels in simplicity, accuracy, and speed; it is easy to implement and fast to run, thus it can be applied to antivirus software, smart firewalls, web applications, etc. [ABSTRACT FROM AUTHOR]

In the distance estimation scheme using Frequency-Modulated-Continuous-Wave (FMCW) radar, the frequency difference, which was caused by the time delay of the received signal reflected from the target, is calculated to estimate the distance information of the target. In this paper, we propose a distance estimation scheme exploiting the deep learning technology of artificial neural network to improve the accuracy of distance estimation over the conventional Fast Fourier Transform (FFT) Max value index-based distance estimation scheme. The performance of the proposed scheme is compared with that of the conventional scheme through the experiments evaluating the accuracy of distance estimation. The average estimated distance error of the proposed scheme was 0.069 m, while that of the conventional scheme was 1.9 m. [ABSTRACT FROM AUTHOR]

Out-of-vocabulary (OOV) words are the most challenging problem in automatic speech recognition (ASR), especially for morphologically rich languages. Most end-to-end speech recognition systems are performed at word and character levels of a language. Amharic is a poorly resourced but morphologically rich language. This paper proposes hybrid connectionist temporal classification with attention end-to-end architecture and a syllabification algorithm for Amharic automatic speech recognition system (AASR) using its phoneme-based subword units. This algorithm helps to insert the epithetic vowel እ[ɨ], which is not included in our Grapheme-to-Phoneme (G2P) conversion algorithm developed using consonant–vowel (CV) representations of Amharic graphemes. The proposed end-to-end model was trained in various Amharic subwords, namely characters, phonemes, character-based subwords, and phoneme-based subwords generated by the byte-pair-encoding (BPE) segmentation algorithm. Experimental results showed that context-dependent phoneme-based subwords tend to result in more accurate speech recognition systems than the character-based, phoneme-based, and character-based subword counterparts. Further improvement was also obtained in proposed phoneme-based subwords with the syllabification algorithm and SpecAugment data augmentation technique. The word error rate (WER) reduction was 18.38% compared to character-based acoustic modeling with the word-based recurrent neural network language modeling (RNNLM) baseline. These phoneme-based subword models are also useful to improve machine and speech translation tasks. [ABSTRACT FROM AUTHOR]

This paper is devoted to solving the problem of determining the age of the author of the text based on models of deep neural networks. The article presents an analysis of methods for determining the age of the author of a text and approaches to determining the age of a user by a photo. This could be a solution to the problem of inaccurate data for training by filtering out incorrect user-specified age data. A detailed description of the author's technique based on deep neural network models and the interpretation of the results is also presented. The study found that the proposed technique achieved 82% accuracy in determining the age of the author from Russian-language text, which makes it competitive in comparison with approaches for other languages. [ABSTRACT FROM AUTHOR]

Deep Neural Networks are known for impressive results in a wide range of applications, being responsible for many advances in technology over the past few years. However, debugging and understanding neural networks models' inner workings is a complex task, as there are several parameters and variables involved in every decision. Multidimensional projection techniques have been successfully adopted to display neural network hidden layer outputs in an explainable manner, but comparing different outputs often means overlapping projections or observing them side-by-side, presenting hurdles for users in properly conveying data flow. In this paper, we introduce a novel approach for comparing projections obtained from multiple stages in a neural network model and visualizing differences in data perception. Changes among projections are transformed into trajectories that, in turn, generate vector fields used to represent the general flow of information. This representation can then be used to create layouts that highlight new information about abstract structures identified by neural networks. [ABSTRACT FROM AUTHOR]

The paper focuses on what two different types of Recurrent Neural Networks, namely a recurrent Long Short-Term Memory and a recurrent variant of self-organizing memories, a Temporal Self-Organizing Map, can tell us about speakers' learning and processing a set of fully inflected verb forms selected from the top-frequency paradigms of Italian and German. Both architectures, due to the re-entrant layer of temporal connectivity, can develop a strong sensitivity to sequential patterns that are highly attested in the training data. The main goal is to evaluate learning and processing dynamics of verb inflection data in the two neural networks by focusing on the effects of morphological structure on word production and word recognition, as well as on word generalization for untrained verb forms. For both models, results show that production and recognition, as well as generalization, are facilitated for verb forms in regular paradigms. However, the two models are differently influenced by structural effects, with the Temporal Self-Organizing Map more prone to adaptively find a balance between processing issues of learnability and generalization, on the one side, and discriminability on the other side. [ABSTRACT FROM AUTHOR]

The present paper discusses a comparative application of image processing techniques, i.e., Discrete Fourier Transform, K-Means clustering and Artificial Neural Network, for the detection of defects in the industrial context of assembled tires. The used Artificial Neural Network technique is based on Long Short-Term Memory and Fully Connected neural networks. The investigations focus on the monitoring and quality control of defects, which may appear on the external surface of tires after being assembled. Those defects are caused from tires which are not properly assembled to their respective metallic wheel rim, generating deformations and scrapes which are not desired. The proposed image processing techniques are applied on raw high-resolution images, which are acquired by in-line imaging and optical instruments. All the described techniques, i.e., Discrete Fourier Transform, K-Means clustering and Long Short-Term Memory, were able to determine defected and acceptable external tire surfaces. The proposed research is taken in the context of an industrial project which focuses on the development of automated quality control and monitoring methodologies, within the field of Industry 4.0 facilities. The image processing techniques are thus meant to be adopted into production processes, giving a strong support to the in-line quality control phase. [ABSTRACT FROM AUTHOR]

Radio is evolving in a changing digital media ecosystem. Audio-on-demand has shaped the landscape of big unstructured audio data available online. In this paper, a framework for knowledge extraction is introduced, to improve discoverability and enrichment of the provided content. A web application for live radio production and streaming is developed. The application offers typical live mixing and broadcasting functionality, while performing real-time annotation as a background process by logging user operation events. For the needs of a typical radio station, a supervised speaker classification model is trained for the recognition of 24 known speakers. The model is based on a convolutional neural network (CNN) architecture. Since not all speakers are known in radio shows, a CNN-based speaker diarization method is also proposed. The trained model is used for the extraction of fixed-size identity d-vectors. Several clustering algorithms are evaluated, having the d-vectors as input. The supervised speaker recognition model for 24 speakers scores an accuracy of 88.34%, while unsupervised speaker diarization scores a maximum accuracy of 87.22%, as tested on an audio file with speech segments from three unknown speakers. The results are considered encouraging regarding the applicability of the proposed methodology. [ABSTRACT FROM AUTHOR]

An accurate estimate of rainfall levels is fundamental in numerous application scenarios: weather forecasting, climate models, design of hydraulic structures, precision agriculture, etc. An accurate estimate becomes essential to be able to warn of the imminent occurrence of a calamitous event and reduce the risk to human beings. Unfortunately, to date, traditional techniques for estimating rainfall levels present numerous critical issues. The algorithm applies the Convolution Neural Network (CNN) directly to the audio signal, using 3 s sliding windows with an offset of only 100 milliseconds. Therefore, by using low cost and low power hardware, the proposed algorithm allows implementing critical high rainfall event alerting mechanisms with short response times and low estimation errors. More specifically, this paper proposes a new approach to rainfall estimation based on the classification of different acoustic timbres that rain produces at different intensities and on CNN. The results obtained on seven classes ranging from "No rain" to "Cloudburst" indicate an average accuracy of 75%, which rises to 93% if the misclassifications of the adjacent classes are not considered. Some application contexts concern smart cities for which the integration of an audio sensor inside the luminaire of a street lamp is foreseen, precision agriculture, as well as highway safety, by minimizing the risks of aquaplaning. [ABSTRACT FROM AUTHOR]

Diseases from Ginkgo biloba have brought great losses to medicine and the economy. Therefore, if the degree of disease can be automatically identified in Ginkgo biloba leaves, people will take appropriate measures to avoid losses in advance. Deep learning has made great achievements in plant disease identification and classification. For this paper, the convolution neural network model was used to classify the different degrees of ginkgo leaf disease. This study used the VGGNet-16 and Inception V3 models. After preprocessing and training 1322 original images under laboratory conditions and 2408 original images under field conditions, 98.44% accuracy was achieved under laboratory conditions and 92.19% under field conditions with the VGG model. The Inception V3 model achieved 92.3% accuracy under laboratory conditions and 93.2% under field conditions. Thus, the Inception V3 model structure was more suitable for field conditions. To our knowledge, there is very little research on the classification of different degrees of the same plant disease. The success of this study will have a significant impact on the prediction and early prevention of ginkgo leaf blight. [ABSTRACT FROM AUTHOR]