Detection and localization of Goliath grouper using their low-frequency pulse sounds.

The goal of this paper is to implement and deploy an automated detector and localization model to locate underwater marine organisms using their low-frequency pulse sounds. This model is based on time difference of arrival (TDOA) and uses a two-stage approach, first, to identify the sound and, second, to localize it. In the first stage, an adaptive matched filter (MF) is designed and implemented to detect and determine the timing of the sound pulses recorded by the hydrophones. The adaptive MF measures the signal and noise levels to determine an adaptive threshold for the pulse detection. In the second stage, the detected sound pulses are fed to a TDOA localization algorithm to compute the locations of the sound source. Despite the uncertainties stemming from various factors that might cause errors in position estimates, it is shown that the errors in source locations are within the dimensions of the array. Further, our method was applied to the localization of Goliath grouper pulse-like calls from a six-hydrophone array. It was revealed that the intrinsic error of the model was about 2 m for an array spanned over 50 m. This method can be used to automatically process large amount of acoustic data and provide a precise description of small scale movements of marine organisms that produce low-frequency sound pulses.

Measurements, Analysis, Classification, and Detection of Gunshot and Gunshot-like Sounds.

Gun violence has been on the rise in recent years. To help curb the downward spiral of this negative influence in communities, machine learning strategies on gunshot detection can be developed and deployed. After outlining the procedure by which a typical type of gunshot-like sounds were measured, this paper focuses on the analysis of feature importance pertaining to gunshot and gunshot-like sounds. The random forest mean decrease in impurity and the SHapley Additive exPlanations feature importance analysis were employed for this task. From the feature importance analysis, feature reduction was then carried out. Via the Mel-frequency cepstral coefficients feature extraction process on 1-sec audio clips, these extracted features were then reduced to a more manageable quantity using the above-mentioned feature reduction processes. These reduced features were sent to a random forest classifier. The SHapley Additive exPlanations feature importance output was compared to that of the mean decrease in impurity feature importance. The results show what Mel-frequency cepstral coefficients features are important in discriminating gunshot sounds and various gunshot-like sounds. Together with the feature importance/reduction processes, the recent uniform manifold approximation and projection method was used to compare the closeness of various gunshot-like sounds to gunshot sounds in the feature space. Finally, the approach presented in this paper provides people with a viable means to make gunshot sounds more discernible from other sounds.

A multimodel deep learning algorithm to detect North Atlantic right whale up-calls.

We present a new method of detecting North Atlantic Right Whale (NARW) upcalls using a Multimodel Deep Learning (MMDL) algorithm. A MMDL detector is a classifier that embodies Convolutional Neural Networks (CNNs) and Stacked Auto Encoders (SAEs) and a fusion classifier to evaluate their output for a final decision. The MMDL detector aims for diversity in the choice of the classifier so that its architecture is learned to fit the data. Spectrograms and scalograms of signals from passive acoustic sensors are used to train the MMDL detector. Guided by previous applications, we trained CNNs with spectrograms and SAEs with scalograms. Outputs from individual models were evaluated by the fusion classifier. The results obtained from the MMDL algorithm were compared to those obtained from conventional machine learning algorithms trained with handcrafted features. It showed the superiority of the MMDL algorithm in terms of the upcall detection rate, non-upcall detection rate, and false alarm rate. The autonomy of the MMDL detector has immediate application to the effective monitoring and protection of one of the most endangered species in the world where accurate call detection of a low-density species is critical, especially in environments of high acoustic-masking.

Data Collection, Modeling, and Classification for Gunshot and Gunshot-like Audio Events: A Case Study.

Distinguishing between a dangerous audio event like a gun firing and other non-life-threatening events, such as a plastic bag bursting, can mean the difference between life and death and, therefore, the necessary and unnecessary deployment of public safety personnel. Sounds generated by plastic bag explosions are often confused with real gunshot sounds, by either humans or computer algorithms. As a case study, the research reported in this paper offers insight into sounds of plastic bag explosions and gunshots. An experimental study in this research reveals that a deep learning-based classification model trained with a popular urban sound dataset containing gunshot sounds cannot distinguish plastic bag pop sounds from gunshot sounds. This study further shows that the same deep learning model, if trained with a dataset containing plastic pop sounds, can effectively detect the non-life-threatening sounds. For this purpose, first, a collection of plastic bag-popping sounds was recorded in different environments with varying parameters, such as plastic bag size and distance from the recording microphones. The audio clips' duration ranged from 400 ms to 600 ms. This collection of data was then used, together with a gunshot sound dataset, to train a classification model based on a convolutional neural network (CNN) to differentiate life-threatening gunshot events from non-life-threatening plastic bag explosion events. A comparison between two feature extraction methods, the Mel-frequency cepstral coefficients (MFCC) and Mel-spectrograms, was also done. Experimental studies conducted in this research show that once the plastic bag pop sounds are injected into model training, the CNN classification model performs well in distinguishing actual gunshot sounds from plastic bag sounds.

Classifying Goliath Grouper (Epinephelus itajara) Behaviors from a Novel, Multi-Sensor Tag.

Inertial measurement unit sensors (IMU; i.e., accelerometer, gyroscope and magnetometer combinations) are frequently fitted to animals to better understand their activity patterns and energy expenditure. Capable of recording hundreds of data points a second, these sensors can quickly produce large datasets that require methods to automate behavioral classification. Here, we describe behaviors derived from a custom-built multi-sensor bio-logging tag attached to Atlantic Goliath grouper (Epinephelus itajara) within a simulated ecosystem. We then compared the performance of two commonly applied machine learning approaches (random forest and support vector machine) to a deep learning approach (convolutional neural network, or CNN) for classifying IMU data from this tag. CNNs are frequently used to recognize activities from IMU data obtained from humans but are less commonly considered for other animals. Thirteen behavioral classes were identified during ethogram development, nine of which were classified. For the conventional machine learning approaches, 187 summary statistics were extracted from the data, including time and frequency domain features. The CNN was fed absolute values obtained from fast Fourier transformations of the raw tri-axial accelerometer, gyroscope and magnetometer channels, with a frequency resolution of 512 data points. Five metrics were used to assess classifier performance; the deep learning approach performed better across all metrics (Sensitivity = 0.962; Specificity = 0.996; F1-score = 0.962; Matthew's Correlation Coefficient = 0.959; Cohen's Kappa = 0.833) than both conventional machine learning approaches. Generally, the random forest performed better than the support vector machine. In some instances, a conventional learning approach yielded a higher performance metric for particular classes (e.g., the random forest had a F1-score of 0.971 for backward swimming compared to 0.955 for the CNN). Deep learning approaches could potentially improve behavioral classification from IMU data, beyond that obtained from conventional machine learning methods.

Transfer learning for efficient classification of grouper sound.

A transfer learning approach is proposed to classify grouper species by their courtship-associated sounds produced during spawning aggregations. Vessel sounds are also included in order to potentially identify human interaction with spawning fish. Grouper sounds recorded during spawning aggregations were first converted to time-frequency representations. Two types of time frequency representations were used in this study: spectrograms and scalograms. These were converted to images, and then fed to pretrained deep neural network models: VGG16, VGG19, Google Net, and MobileNet. The experimental results revealed that transfer learning significantly outperformed the manually identified features approach for grouper sound classification. In addition, both time-frequency representations produced almost identical results in terms of classification accuracy.

Classification of red hind grouper call types using random ensemble of stacked autoencoders.

In this paper, a method is introduced for the classification of call types of red hind grouper, an important fishery resource in the Caribbean that produces sounds associated with reproductive behaviors during yearly spawning aggregations. For the undertaken task, two distinct call types of red hind are analyzed. An ensemble of stacked autoencoders (SAEs) is then designed by randomly selecting the hyperparameters of SAEs in the network. These hyperparameters include a number of hidden layers in each SAE and a number of nodes in each hidden layer. Spectrograms of red hind calls are used to train this randomly generated ensemble of SAEs one at a time. Once all individual SAEs are trained, this ensemble is used as a whole to classify call types of red hind. More specifically, the outputs of individual SAEs are combined with a fusion mechanism to produce a final decision on the call type of the input red hind sound. Experimental results show that the innovative approach produces superior results in comparison with those obtained by non-ensemble methods. The algorithm reliably classified red hind call types with over 90% accuracy and successfully detected some calls missed by human observers.

Automatic classification of grouper species by their sounds using deep neural networks.

In this paper, the effectiveness of deep learning for automatic classification of grouper species by their vocalizations has been investigated. In the proposed approach, wavelet denoising is used to reduce ambient ocean noise, and a deep neural network is then used to classify sounds generated by different species of groupers. Experimental results for four species of groupers show that the proposed approach achieves a classification accuracy of around 90% or above in all of the tested cases, a result that is significantly better than the one obtained by a previously reported method for automatic classification of grouper calls.

An approach for automatic classification of grouper vocalizations with passive acoustic monitoring.

Grouper, a family of marine fishes, produce distinct vocalizations associated with their reproductive behavior during spawning aggregation. These low frequencies sounds (50-350 Hz) consist of a series of pulses repeated at a variable rate. In this paper, an approach is presented for automatic classification of grouper vocalizations from ambient sounds recorded in situ with fixed hydrophones based on weighted features and sparse classifier. Group sounds were labeled initially by humans for training and testing various feature extraction and classification methods. In the feature extraction phase, four types of features were used to extract features of sounds produced by groupers. Once the sound features were extracted, three types of representative classifiers were applied to categorize the species that produced these sounds. Experimental results showed that the overall percentage of identification using the best combination of the selected feature extractor weighted mel frequency cepstral coefficients and sparse classifier achieved 82.7% accuracy. The proposed algorithm has been implemented in an autonomous platform (wave glider) for real-time detection and classification of group vocalizations.

Identification of Gene Expression in Different Stages of Breast Cancer with Machine Learning.

Determining the tumor origin in humans is vital in clinical applications of molecular diagnostics. Metastatic cancer is usually a very aggressive disease with limited diagnostic procedures, despite the fact that many protocols have been evaluated for their effectiveness in prognostication. Research has shown that dysregulation in miRNAs (a class of non-coding, regulatory RNAs) is remarkably involved in oncogenic conditions. This research paper aims to develop a machine learning model that processes an array of miRNAs in 1097 metastatic tissue samples from patients who suffered from various stages of breast cancer. The suggested machine learning model is fed with miRNA quantitative read count data taken from The Cancer Genome Atlas Data Repository. Two main feature-selection techniques have been used, mainly Neighborhood Component Analysis and Minimum Redundancy Maximum Relevance, to identify the most discriminant and relevant miRNAs for their up-regulated and down-regulated states. These miRNAs are then validated as biological identifiers for each of the four cancer stages in breast tumors. Both machine learning algorithms yield performance scores that are significantly higher than the traditional fold-change approach, particularly in earlier stages of cancer, with Neighborhood Component Analysis and Minimum Redundancy Maximum Relevance achieving accuracy scores of up to 0.983 and 0.931, respectively, compared to 0.920 for the FC method. This study underscores the potential of advanced feature-selection methods in enhancing the accuracy of cancer stage identification, paving the way for improved diagnostic and therapeutic strategies in oncology.

Deep learning prediction of two-dimensional ocean dynamics with wavelet-compressed data.

This study addresses the challenge represented by the application of deep learning models to the prediction of ocean dynamics using datasets over a large region or with high spatial or temporal resolution In a previous study by the authors of this article, they showed that such a challenge could be met by using a divide and conquer approach. The domain was in fact split into multiple sub-regions, which were small enough to be predicted individually and in parallel with each other by a deep learning model. At each time step of the prediction process, the sub-model solutions would be merged at the boundary of each sub-region to remove discontinuities between consecutive domains in order to predict the evolution of the full domain. This approach led to the growth of non-dynamical errors that decreased the prediction skill of our model. In the study herein, we show that wavelets can be used to compress the data and reduce its dimension. Each compression level reduces by a factor of two the horizontal resolution of the dataset. We show that despite the loss of information, a level 3 compression produces an improved prediction of the ocean two-dimensional data in comparison to the divide and conquer approach. Our method is evaluated on the prediction of the sea surface height of the most energetic feature of the Gulf of Mexico, namely the Loop Current.

Physics-Informed Tensor-Train ConvLSTM for Volumetric Velocity Forecasting of the Loop Current.

According to the National Academies, a week long forecast of velocity, vertical structure, and duration of the Loop Current (LC) and its eddies at a given location is a critical step toward understanding their effects on the gulf ecosystems as well as toward anticipating and mitigating the outcomes of anthropogenic and natural disasters in the Gulf of Mexico (GoM). However, creating such a forecast has remained a challenging problem since LC behavior is dominated by dynamic processes across multiple time and spatial scales not resolved at once by conventional numerical models. In this paper, building on the foundation of spatiotemporal predictive learning in video prediction, we develop a physics informed deep learning based prediction model called-Physics-informed Tensor-train ConvLSTM (PITT-ConvLSTM)-for forecasting 3D geo-spatiotemporal sequences. Specifically, we propose (1) a novel 4D higher-order recurrent neural network with empirical orthogonal function analysis to capture the hidden uncorrelated patterns of each hierarchy, (2) a convolutional tensor-train decomposition to capture higher-order space-time correlations, and (3) a mechanism that incorporates prior physics from domain experts by informing the learning in latent space. The advantage of our proposed approach is clear: constrained by the law of physics, the prediction model simultaneously learns good representations for frame dependencies (both short-term and long-term high-level dependency) and inter-hierarchical relations within each time frame. Experiments on geo-spatiotemporal data collected from the GoM demonstrate that the PITT-ConvLSTM model can successfully forecast the volumetric velocity of the LC and its eddies for a period greater than 1 week.

Diabetic Retinopathy (DR) Severity Level Classification Using Multimodel Convolutional Neural Networks.

Diabetic retinopathy (DR) is a progressive eye disease that affects a large portion of working-age adults. DR, which may progress to an irreversible state that causes blindness, can be diagnosed with a comprehensive dilated eye exam. With the eye dilated, the Doctor takes pictures of the inside of the eye via a medical procedure called Fluorescein Angiography, in which a dye is injected into the bloodstream. The dye highlights the blood vessels in the back of the eye so they can be photographed. In addition, the Doctor may request an Optical Coherence Tomography (OCT) exam, by which cross-sectional photos of the retina are produced to measure the thickness of the retina. Early prognostication is vital in treating the disease and preventing it from progressing into advanced irreversible stages. Skilled medical personnel and necessary medical facilities are required to detect DR in its five major stages. In this paper, we propose a diagnostic tool to detect Diabetic retinopathy from fundus images by using an ensemble of multi-inception CNN networks. Our inception block consists of three Convolutional layers with kernel sizes of 3x3, 5x5, and 1x1 that are concatenated deeply and forwarded to the max-pooling layer. We experimentally compare our proposed method with two pre-trained models: VGG16 and GoogleNets. The experiment results show that the proposed method can achieve an accuracy of 93.2% by an ensemble of 10 random networks, compared to 81% obtained with transfer learning based on VGG19.

Validation of miRNAs as Breast Cancer Biomarkers with a Machine Learning Approach.

Certain small noncoding microRNAs (miRNAs) are differentially expressed in normal tissues and cancers, which makes them great candidates for biomarkers for cancer. Previously, a selected subset of miRNAs has been experimentally verified to be linked to breast cancer. In this paper, we validated the importance of these miRNAs using a machine learning approach on miRNA expression data. We performed feature selection, using Information Gain (IG), Chi-Squared (CHI2) and Least Absolute Shrinkage and Selection Operation (LASSO), on the set of these relevant miRNAs to rank them by importance. We then performed cancer classification using these miRNAs as features using Random Forest (RF) and Support Vector Machine (SVM) classifiers. Our results demonstrated that the miRNAs ranked higher by our analysis had higher classifier performance. Performance becomes lower as the rank of the miRNA decreases, confirming that these miRNAs had different degrees of importance as biomarkers. Furthermore, we discovered that using a minimum of three miRNAs as biomarkers for breast cancers can be as effective as using the entire set of 1800 miRNAs. This work suggests that machine learning is a useful tool for functional studies of miRNAs for cancer detection and diagnosis.

From Pauling's abzyme concept to the new era of hydrolytic anti-DNA autoantibodies: a link to rational vaccine design? - A review.

Specific entities of naturally-occurring DNA hydrolytic/cytotoxic antibodies (abzymes) are linked to autoimmune and lymphoproliferative disorders. Suggested sequence of underlying activities conform to such entities penetrating the living cells, trans-locating to nucleus and recognising specific binding sites within single- or double-stranded DNA. Their origin is unknown since corresponding immunogens are unidentified. These anti-DNA antibodies could be the organism's immune response to microbial attack. Their structure, function and pathogenicity were investigated in wet-lab and via bioinformatics in context of Rational Vaccine Designs. This paper offers a comprehensive critical review on the subject in the light of known and newly proposed concepts.