Search

We show that machine learning methods, in particular, extreme trees and neural networks (NNs), provide strong statistical evidence in favor of bond return predictability. NN forecasts based on macroeconomic and yield information translate into economic gains that are larger than those obtained using yields alone. Interestingly, the nature of unspanned factors changes along the yield curve: stock- and labor-market-related variables are more relevant for short-term maturities, whereas output and income variables matter more for longer maturities. Finally, NN forecasts correlate with proxies for time-varying risk aversion and uncertainty, lending support to models featuring both channels. [ABSTRACT FROM AUTHOR]

With rapidly evolving high-throughput technologies and consistently decreasing costs, collecting multimodal omics data in large-scale studies has become feasible. Although studying multiomics provides a new comprehensive approach in understanding the complex biological mechanisms of human diseases, the high dimensionality of omics data and the complexity of the interactions among various omics levels in contributing to disease phenotypes present tremendous analytical challenges. There is a great need of novel analytical methods to address these challenges and to facilitate multiomics analyses. In this paper, we propose a multimodal functional deep learning (MFDL) method for the analysis of high-dimensional multiomics data. The MFDL method models the complex relationships between multiomics variants and disease phenotypes through the hierarchical structure of deep neural networks and handles high-dimensional omics data using the functional data analysis technique. Furthermore, MFDL leverages the structure of the multimodal model to capture interactions between different types of omics data. Through simulation studies and real-data applications, we demonstrate the advantages of MFDL in terms of prediction accuracy and its robustness to the high dimensionality and noise within the data. [ABSTRACT FROM AUTHOR]

This study developed a noise-reduction method for acoustic logging array signals using a deep neural network algorithm in the time-frequency domain. Initially, we derived analytical solutions for the received waveforms when the acoustic logging tool was positioned either at the centre or eccentrically within the borehole. To simulate the received waveforms across various formations, we developed a real-axis integration algorithm. Subsequently, we devised a noise-reduction algorithm workflow based on a convolutional neural network and configured the structure and parameters of the U-net using TensorFlow. To address the scarcity of open datasets, we established both signal and noise datasets. The signal dataset was generated using theoretical simulation encompassing various model parameters, while the noise dataset was collected during tool testing and downhole operations. The trained model demonstrated substantial noise-reduction capabilities during validation. To validate the effectiveness of the algorithm, we applied noise reduction to actual data collected during downhole operations in a TangGu oil field, yielding impressive results across different types of noisy data. Therefore, the U-net-based time-domain noise-reduction algorithm proposed in this paper holds the potential to significantly improve the quality of acoustic logging array signals. [ABSTRACT FROM AUTHOR]

This paper introduces a sequential approximate optimization method that combines the finite element method (FEM), dynamic differential evolution (DDE), and artificial neural network (ANN) surrogate models. The developed method is applied to address two optimization problems. The first involves metamaterial design optimization for metal sheet architecture with binary design variables. The second pertains to optimizing process parameters in multi-stage metal forming, where the discrete nature arises owing to changing tool geometries across stages. This process is highly non-linear, accumulating contact, geometric, and material non-linear effects discretely through forming stages. The efficacy of the proposed optimization method, utilizing ANN surrogate models, is compared with traditionally used polynomial response surface (PRS) surrogate models, primarily based on low-order polynomials. Efficient learning of ANN surrogate models is facilitated through the FEM and Python integration framework. Initial data for surrogate model training are collected via Latin hypercube sampling and FEM simulations. DDE is employed for sequential approximate optimization, optimizing ANN or PRS surrogate models to determine optimal design variables. PRS surrogate models encounter challenges in dealing with non-linear changes in sequential approximate optimization concerning discrete characteristics such as binary design variables and discrete non-linear behavior found in multi-stage metal forming processes. Owing to the discrete nature, PRS surrogate models require more data and iterations for optimal design variables. In contrast, ANN surrogate models adeptly predict non-linear behavior through the activation function's characteristics. In the optimization problem of metal sheet architecture for design Target C, the ANN surrogate model required an average of 4.6 times fewer iterations to satisfy stopping criteria compared with the PRS surrogate model. Furthermore, in the optimization of multi-stage deep drawing processes, the ANN surrogate model required an average of 6.1 times fewer iterations to satisfy stopping criteria compared with the PRS surrogate model. As a result, the sequential global optimization method utilizing ANN surrogate models achieves optimal design variables with fewer iterations than PRS surrogate models. Further confirmation of the method's efficiency is provided by comparing Pearson correlation coefficients and locus plots. [ABSTRACT FROM AUTHOR]

This paper looks at the representation of supra-classical, non-monotonic (SCNM) logic by an artificial neural network. It identifies the features of defeasiblity in this logic related to inference in the context of common-sense reasoning. It considers the machine characteristics that make a representation possible, with reference to previous literature. We describe a theoretical environment for investigating the representation and provide experimental evidence confirming that a Boltzmann machine is a suitable network representation. A Boltzmann machine can learn an input distribution corresponding to a preference relation and explicitly retrieve appropriate model states, constituting one-to-many mappings, entailed by the uncertain information contained in a premiss. The place of the Boltzmann machine in knowledge representation is discussed. In future papers, this neural network model of SCNM logic will serve as an experimental gateway for exploration of typicality and belief revision. [ABSTRACT FROM AUTHOR]

The nine papers included in this special issue represent a selection of extended contributions presented at the 14th International Conference on Hybrid Artificial Intelligent Systems (HAIS), HAIS 2019, held in León, Spain, 4-6 June 2019, and organized by the BISITE group and the University of León. In order to solve this complex optimization problem, authors propose an efficient hybrid algorithm, which was obtained by incorporating a linear programming optimization problem within the framework of a genetic algorithm. The aim of the next research, by Simic I et al. i , is the classification of skin tumours by analysing medical skin tumour dermoscopy images. [Extracted from the article]

Spiking neural networks (SNNs) are gaining increasing attention for their biological plausibility and potential for improved computational efficiency. To match the high spatial-temporal dynamics in SNNs, neuromorphic chips are highly desired to execute SNNs in hardware-based neuron and synapse circuits directly. This paper presents a large-scale neuromorphic chip named Darwin3 with a novel instruction set architecture, which comprises 10 primary instructions and a few extended instructions. It supports flexible neuron model programming and local learning rule designs. The Darwin3 chip architecture is designed in a mesh of computing nodes with an innovative routing algorithm. We used a compression mechanism to represent synaptic connections, significantly reducing memory usage. The Darwin3 chip supports up to 2.35 million neurons, making it the largest of its kind on the neuron scale. The experimental results showed that the code density was improved by up to 28.3× in Darwin3, and that the neuron core fan-in and fan-out were improved by up to 4096× and 3072× by connection compression compared to the physical memory depth. Our Darwin3 chip also provided memory saving between 6.8× and 200.8× when mapping convolutional spiking neural networks onto the chip, demonstrating state-of-the-art performance in accuracy and latency compared to other neuromorphic chips. [ABSTRACT FROM AUTHOR]

Brain-inspired computing, drawing inspiration from the fundamental structure and information-processing mechanisms of the human brain, has gained significant momentum in recent years. It has emerged as a research paradigm centered on brain–computer dual-driven and multi-network integration. One noteworthy instance of this paradigm is the hybrid neural network (HNN), which integrates computer-science-oriented artificial neural networks (ANNs) with neuroscience-oriented spiking neural networks (SNNs). HNNs exhibit distinct advantages in various intelligent tasks, including perception, cognition and learning. This paper presents a comprehensive review of HNNs with an emphasis on their origin, concepts, biological perspective, construction framework and supporting systems. Furthermore, insights and suggestions for potential research directions are provided aiming to propel the advancement of the HNN paradigm. [ABSTRACT FROM AUTHOR]

Background Risk evaluation of lymph node metastasis for early-stage (T1 and T2) colorectal cancers is critical for determining therapeutic strategies. Traditional methods of lymph node metastasis prediction have limited accuracy. This systematic review aimed to review the potential of artificial intelligence in predicting lymph node metastasis in early-stage colorectal cancers. Methods A comprehensive search was performed of papers that evaluated the potential of artificial intelligence in predicting lymph node metastasis in early-stage colorectal cancers. Studies were appraised using the Joanna Briggs Institute tools. The primary outcome was summarizing artificial intelligence models and their accuracy. Secondary outcomes included influential variables and strategies to address challenges. Results Of 3190 screened manuscripts, 11 were included, involving 8648 patients from 1996 to 2023. Due to diverse artificial intelligence models and varied metrics, no data synthesis was performed. Models included random forest algorithms, support vector machine, deep learning, artificial neural network, convolutional neural network and least absolute shrinkage and selection operator regression. Artificial intelligence models' area under the curve values ranged from 0.74 to 0.9993 (slide level) and 0.9476 to 0.9956 (single-node level), outperforming traditional clinical guidelines. Conclusion Artificial intelligence models show promise in predicting lymph node metastasis in early-stage colorectal cancers, potentially refining clinical decisions and improving outcomes. PROSPERO registration number CRD42023409094. [ABSTRACT FROM AUTHOR]

Connections between circular RNAs (circRNAs) and microRNAs (miRNAs) assume a pivotal position in the onset, evolution, diagnosis and treatment of diseases and tumors. Selecting the most potential circRNA-related miRNAs and taking advantage of them as the biological markers or drug targets could be conducive to dealing with complex human diseases through preventive strategies, diagnostic procedures and therapeutic approaches. Compared to traditional biological experiments, leveraging computational models to integrate diverse biological data in order to infer potential associations proves to be a more efficient and cost-effective approach. This paper developed a model of Convolutional Autoencoder for CircRNA–MiRNA Associations (CA-CMA) prediction. Initially, this model merged the natural language characteristics of the circRNA and miRNA sequence with the features of circRNA–miRNA interactions. Subsequently, it utilized all circRNA–miRNA pairs to construct a molecular association network, which was then fine-tuned by labeled samples to optimize the network parameters. Finally, the prediction outcome is obtained by utilizing the deep neural networks classifier. This model innovatively combines the likelihood objective that preserves the neighborhood through optimization, to learn the continuous feature representation of words and preserve the spatial information of two-dimensional signals. During the process of 5-fold cross-validation, CA-CMA exhibited exceptional performance compared to numerous prior computational approaches, as evidenced by its mean area under the receiver operating characteristic curve of 0.9138 and a minimal SD of 0.0024. Furthermore, recent literature has confirmed the accuracy of 25 out of the top 30 circRNA–miRNA pairs identified with the highest CA-CMA scores during case studies. The results of these experiments highlight the robustness and versatility of our model. [ABSTRACT FROM AUTHOR]

Aspect-based sentiment analysis aims to identify the sentiment polarity of aspects in a given sentence. Although existing neural network models show promising results, they cannot meet the expectations in the case of a single network structure and limited dataset. When an aspect term composes more than one word, many models use the coarse-grained attention mechanism but lead to the unsatisfactory results. Besides, the relative distance between words in a sentence is always out of consideration. In this paper, we propose a model based on the interaction matrix and global attention mechanism to improve the ability of aspect-based sentiment analysis. First of all, the relative distance features of words in a sentence are initialized to enrich word embedding. Second, classic neural networks are applied to extract the essential features of word embedding in a sentence, such as long short-term memory and convolutional neural network. Third, an interaction matrix and global attention mechanism are combined to calculate weighted scores and measure relationships between aspect terms and context words. Finally, sentiment polarity is represented through a softmax layer. Experimental results on restaurant, laptop and twitter datasets show that the performance of the proposed model is superior to other methods. [ABSTRACT FROM AUTHOR]

In recent years, deep neural networks have been extensively applied in various fields, and face recognition is one of the most important applications. Artificial intelligence has reached or even surpassed human capabilities in many fields. However, while artificial intelligence application provides convenience to the human lives, it also leads to the risk of privacy leaking. At present, the privacy protection technology for human faces has received extensive attention. Research goals of face privacy protection technology mainly include providing face anonymization and data availability protection. Existing methods usually have insufficient anonymity and they are not easy to control the degree of image distortion, which makes it difficult to achieve the purpose of privacy protection. Moreover, they do not explicitly perform diversity preservation of attributes such as emotions, expressions and ethnicities, so they cannot perform data analysis tasks on non-identity attributes. This paper proposes a diverse privacy face image generation algorithm based on machine learning, called DIVFGEN. This algorithm comprehensively considers image distortion, identity mapping distance loss and emotion classification loss; transforms the privacy protection target into the problem of generating adversarial examples based on the recognition model; and uses an adaptive optimization algorithm to generate anonymity and diversity of privacy images. The experimental results show that on the Cohn-Kanade+ dataset, our algorithm can reduce the probability of facial recognition by the neural network when it accurately classifies sentiment, from 98.6% to 4.8%. [ABSTRACT FROM AUTHOR]

Predicting the remaining useful life (RUL) of the aircraft engine based on historical data plays a pivotal role in formulating maintenance strategies and mitigating the risk of critical failures. None the less, attaining precise RUL predictions often encounters challenges due to the scarcity of historical condition monitoring data. This paper introduces a multiscale deep transfer learning framework via integrating domain adaptation principles. The framework encompasses three integral components: a feature extraction module, an encoding module, and an RUL prediction module. During pre-training phase, the framework leverages a multiscale convolutional neural network to extract distinctive features from data across varying scales. The ensuing parameter transfer adopts a domain adaptation strategy centered around maximum mean discrepancy. This method efficiently facilitates the acquisition of domain-invariant features from the source and target domains. The refined domain adaptation Transformer-based multiscale convolutional neural network model exhibits enhanced suitability for predicting RUL in the target domain under the condition of limited samples. Experiments on the C-MAPSS dataset have shown that the proposed method significantly outperforms state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Computational progressive failure analysis of carbon fiber reinforced polymer composite is of vital importance in the verification and validation process of the structural integrity and damage tolerance of modern lightweight aeronautical structures. Enhanced Schapery theory (EST) has been developed and applied to predict the damage pattern and load-bearing capacity of various composite structures. In this paper, EST is enhanced by a deep neural network (DNN) model, which enables fast and accurate predictions of matrix cracking angles under arbitrary stress states of any composite laminate. The DNN model is trained by TensorFlow based on data generated by a damage initiation criterion, which originates from the Mohr–Coulomb failure theory. The EST-DNN model is applied to open-hole tension/compression problems. The results from the EST-DNN model are obtained with no loss in accuracy. The results presented combine the efficient and accurate predicting capabilities brought by machine learning tools and the robustness and user-friendliness of the EST finite element model. [ABSTRACT FROM AUTHOR]

Detecting intrusions in large networks is a highly demanding task. In order to reduce the computation demand of analysing every single packet travelling along one of such networks, some years ago flows were proposed as a way of summarizing traffic information. Very few research works have addressed intrusion detection in flows from a visualizations perspective. In order to bridge this gap, the present paper proposes the application of a novel projection method (Beta Hebbian Learning) under this framework. With the aim to validate this method, 8 traffic segments, containing many flows, have been analysed by means of this projection method. The promising results obtained for these segments, extracted from the University of Twente dataset, validate the proposed application. [ABSTRACT FROM AUTHOR]

This section presents abstracts on papers about logic, including constraint diagrams, neural networks, and dialogue games for argumentation.

In the absence of sufficient user behavior data, game recommendation systems face the cold-start problem. To address this issue, this paper proposes a solution based on the Graph Neural Network pre-training model to alleviate the cold-start problem. The proposed model directly reconstructs cold-start user/game embeddings using a meta-learning setup based on dataset training simulations and uses an adaptive neighbor sampler to improve user interaction relations and thereby to improve game recommendation performance. Experimental results demonstrate the effectiveness and practicality of the recommendation model proposed in this study. Moreover, the proposed model is embedded in the game recommendation system to visualize the recommendation results. [ABSTRACT FROM AUTHOR]

Overparameterized models may have many interpolating solutions; implicit regularization refers to the hidden preference of a particular optimization method towards a certain interpolating solution among the many. A by now established line of work has shown that (stochastic) gradient descent tends to have an implicit bias towards low rank and/or sparse solutions when used to train deep linear networks, explaining to some extent why overparameterized neural network models trained by gradient descent tend to have good generalization performance in practice. However, existing theory for square-loss objectives often requires very small initialization of the trainable weights, which is at odds with the larger scale at which weights are initialized in practice for faster convergence and better generalization performance. In this paper, we aim to close this gap by incorporating and analysing gradient flow (continuous-time version of gradient descent) with weight normalization , where the weight vector is reparameterized in terms of polar coordinates, and gradient flow is applied to the polar coordinates. By analysing key invariants of the gradient flow and using Lojasiewicz's Theorem, we show that weight normalization also has an implicit bias towards sparse solutions in the diagonal linear model, but that in contrast to plain gradient flow, weight normalization enables a robust bias that persists even if the weights are initialized at practically large scale. Experiments suggest that the gains in both convergence speed and robustness of the implicit bias are improved dramatically using weight normalization in overparameterized diagonal linear network models. [ABSTRACT FROM AUTHOR]

This article discusses two papers that address the statistical aspects of machine learning. Both papers utilize neural networks, specifically deep neural networks, to enhance the modeling process. The first paper focuses on automatic changepoint detection in time series using deep learning, while the second paper explores denoising diffusions in non-Euclidean spaces. The authors raise questions about the selection of neural network architecture and the computational cost compared to traditional statistical models. They also highlight the challenges of estimating a large number of parameters in neural networks with small datasets. Overall, these papers contribute to the field of statistical machine learning and suggest avenues for future research. [Extracted from the article]

Artificial neural networks (ANNs), sequence-to-sequence long short-term memory networks (LSTMs) and a newly proposed method combing a rolling window with the latter (LSTM SB NA sb ) are used as the forecasting techniques. In the third contribution, which is by Teso-Fz-Betoño I et al. i , an artificial neural network (ANN) is developed to estimate the relative rotation and translation of the autonomous mobile robot (AMR). In this hybrid algorithm, a genetic algorithm (GA), tour construction and tour improvement algorithms (TCTIAs) and a list-based simulated annealing (LBSA) algorithm are used. [Extracted from the article]

Multi-drug combinations for the treatment of complex diseases are gradually becoming an important treatment, and this type of treatment can take advantage of the synergistic effects among drugs. However, drug–drug interactions (DDIs) are not just all beneficial. Accurate and rapid identifications of the DDIs are essential to enhance the effectiveness of combination therapy and avoid unintended side effects. Traditional DDIs prediction methods use only drug sequence information or drug graph information, which ignores information about the position of atoms and edges in the spatial structure. In this paper, we propose Molormer, a method based on a lightweight attention mechanism for DDIs prediction. Molormer takes the two-dimension (2D) structures of drugs as input and encodes the molecular graph with spatial information. Besides, Molormer uses lightweight-based attention mechanism and self-attention distilling to process spatially the encoded molecular graph, which not only retains the multi-headed attention mechanism but also reduces the computational and storage costs. Finally, we use the Siamese network architecture to serve as the architecture of Molormer, which can make full use of the limited data to train the model for better performance and also limit the differences to some extent between networks dealing with drug features. Experiments show that our proposed method outperforms state-of-the-art methods in Accuracy, Precision, Recall and F1 on multi-label DDIs dataset. In the case study section, we used Molormer to make predictions of new interactions for the drugs Aliskiren, Selexipag and Vorapaxar and validated parts of the predictions. Code and models are available at https://github.com/IsXudongZhang/Molormer. [ABSTRACT FROM AUTHOR]

Birth–death (BD) models are widely used in combination with species phylogenies to study past diversification dynamics. Current inference approaches typically rely on likelihood-based methods. These methods are not generalizable, as a new likelihood formula must be established each time a new model is proposed; for some models, such a formula is not even tractable. Deep learning can bring solutions in such situations, as deep neural networks can be trained to learn the relation between simulations and parameter values as a regression problem. In this paper, we adapt a recently developed deep learning method from pathogen phylodynamics to the case of diversification inference, and we extend its applicability to the case of the inference of state-dependent diversification models from phylogenies associated with trait data. We demonstrate the accuracy and time efficiency of the approach for the time-constant homogeneous BD model and the Binary-State Speciation and Extinction model. Finally, we illustrate the use of the proposed inference machinery by reanalyzing a phylogeny of primates and their associated ecological role as seed dispersers. Deep learning inference provides at least the same accuracy as likelihood-based inference while being faster by several orders of magnitude, offering a promising new inference approach for the deployment of future models in the field. [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) systems utilizing deep neural networks and machine learning (ML) algorithms are widely used for solving critical problems in bioinformatics, biomedical informatics and precision medicine. However, complex ML models that are often perceived as opaque and black-box methods make it difficult to understand the reasoning behind their decisions. This lack of transparency can be a challenge for both end-users and decision-makers, as well as AI developers. In sensitive areas such as healthcare, explainability and accountability are not only desirable properties but also legally required for AI systems that can have a significant impact on human lives. Fairness is another growing concern, as algorithmic decisions should not show bias or discrimination towards certain groups or individuals based on sensitive attributes. Explainable AI (XAI) aims to overcome the opaqueness of black-box models and to provide transparency in how AI systems make decisions. Interpretable ML models can explain how they make predictions and identify factors that influence their outcomes. However, the majority of the state-of-the-art interpretable ML methods are domain-agnostic and have evolved from fields such as computer vision, automated reasoning or statistics, making direct application to bioinformatics problems challenging without customization and domain adaptation. In this paper, we discuss the importance of explainability and algorithmic transparency in the context of bioinformatics. We provide an overview of model-specific and model-agnostic interpretable ML methods and tools and outline their potential limitations. We discuss how existing interpretable ML methods can be customized and fit to bioinformatics research problems. Further, through case studies in bioimaging, cancer genomics and text mining, we demonstrate how XAI methods can improve transparency and decision fairness. Our review aims at providing valuable insights and serving as a starting point for researchers wanting to enhance explainability and decision transparency while solving bioinformatics problems. GitHub: https://github.com/rezacsedu/XAI-for-bioinformatics. [ABSTRACT FROM AUTHOR]

The skin, uniquely positioned at the interface between the human body and the external world, plays a multifaceted immunologic role in human life. In medical practice, early accurate detection of all types of skin tumours is essential to guide appropriate management and improve patients' survival. The most important issue is to differentiate between malignant skin tumours and benign lesions. The aim of this research is the classification of skin tumours by analysing medical skin tumour dermoscopy images. This paper is focused on a new strategy based on deep convolutional neural networks which have recently shown a state-of-the-art performance to define strategy to automatic classification for skin tumour images. The proposed system is tested on well-known HAM10000 data set. For experimental results, verification is performed and the results are compared with similar researches. [ABSTRACT FROM AUTHOR]

In this technological era, human activity recognition (HAR) plays a significant role in several applications like surveillance, health services, Internet of Things, etc. Recent advancements in deep learning and video summarization have motivated us to integrate these techniques for HAR. This paper introduces a computationally efficient HAR technique based on a deep learning framework, which works well in realistic and multi-view environments. Deep convolutional neural networks (DCNNs) normally suffer from different constraints, including data size dependencies, computational complexity, overfitting, training challenges and vanishing gradients. Additionally, with the use of advanced mobile vision devices, the demand for computationally efficient HAR algorithms with the requirement of limited computational resources is high. To address these issues, we used integration of DCNN with video summarization using keyframes. The proposed technique offers a solution that enhances performance with efficient resource utilization. For this, first, we designed a lightweight and computationally efficient deep learning architecture based on the concept of identity skip connections (features reusability), which preserves the gradient loss attenuation and can handle the enormous complexity of activity classes. Subsequently, we employed an efficient keyframe extraction technique to minimize redundancy and succinctly encapsulate the entire video content in a lesser number of frames. To evaluate the efficacy of the proposed method, we performed the experimentation on several publicly available datasets. The performance of the proposed method is measured in terms of evaluation parameters Precision, Recall, F-Measure and Classification Accuracy. The experimental results demonstrated the superiority of the presented algorithm over other existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Website fingerprinting (WF) attacks on Tor enable a passive adversary to predict the encrypted web browsing activity of a victim by matching the eavesdropped traffic with pretrained classifiers. Nowadays, deep learning-based methods have led to significant achievements in the single-tab Tor WF attack. However, the practical implementation of these single-tab methods is challenging due to most real-world Tor traffic involving multiple tabs. Existing single-tab methods hardly identify multi-tab WF due to the overlapping areas of the traffic trace confusing the original features. In this paper, we propose a Trace Image-based Object Detection model named TIOD as a novel multi-tab attacking model. Specifically, we model the traffic overlap in Tor as the object overlap in object detection tasks from the computer vision field. Besides, we propose a special S-matrix scheme to convert a trace into an image: (i) Retaining the original features by keeping the direction and order of the cells and (ii) Bringing cells of the same page closer together in space. We then utilize a specially designed object detection model for trace images, WF R-CNN, to extract features and identify potential destination websites within multi-tab traces. Comparative analysis with other multi-tab attacks is conducted, and the empirical results consistently underscore the superior performance of the proposed trace image model across diverse datasets. In the two-tab setting, TIOD achieves the best accuracy with more than 85% on both the first and second tabs. [ABSTRACT FROM AUTHOR]

Blind image watermarking is regarded as a vital technology to provide copyright of digital images. Due to the rapid growth of deep neural networks, deep learning-based watermarking methods have been widely studied. However, most existing methods which adopt simple embedding and extraction structures cannot fully utilize the image features. In this paper, we propose a novel Single-Encoder-Dual-Decoder (SEDD) watermarking architecture to achieve high imperceptibility and strong robustness. Precisely, the single encoder utilizes normalizing flow to realize watermark embedding, which can effectively fuse the watermark and cover image. For watermark extraction, we introduce a parallel dual-decoder to improve the imperceptibility and extracting ability. Extensive experiments demonstrate that better watermark robustness and imperceptibility are obtained by SEDD architecture. Our method achieves a bit error rate less than 0.1% under most attacks such as JPEG compression, Gaussian blur and crop. Besides, the proposed method also obtains strong robustness under combined attacks and social platform processing. [ABSTRACT FROM AUTHOR]

The prosperity of social IoT data brings revolutionary changes to our daily lives and greatly increases the existing data volume. The edge server trains the local generator and discriminator using the perturbed data, which provides privacy guarantees for the gradient attack on the FedGAN without compromising accuracy. The authors empirically compare the proposed approach with several state-of-the-art anomaly detection methods on real-world datasets, anomaly benchmarks and synthetic datasets. [Extracted from the article]

Nucleic-acid G-quadruplexes (G4s) play vital roles in many cellular processes. Due to their importance, researchers have developed experimental assays to measure nucleic-acid G4s in high throughput. The generated high-throughput datasets gave rise to unique opportunities to develop machine-learning-based methods, and in particular deep neural networks, to predict G4s in any given nucleic-acid sequence and any species. In this paper, we review the success stories of deep-neural-network applications for G4 prediction. We first cover the experimental technologies that generated the most comprehensive nucleic-acid G4 high-throughput datasets in recent years. We then review classic rule-based methods for G4 prediction. We proceed by reviewing the major machine-learning and deep-neural-network applications to nucleic-acid G4 datasets and report a novel comparison between them. Next, we present the interpretability techniques used on the trained neural networks to learn key molecular principles underlying nucleic-acid G4 folding. As a new result, we calculate the overlap between measured DNA and RNA G4s and compare the performance of DNA- and RNA-G4 predictors on RNA- and DNA-G4 datasets, respectively, to demonstrate the potential of transfer learning from DNA G4s to RNA G4s. Last, we conclude with open questions in the field of nucleic-acid G4 prediction and computational modeling. [ABSTRACT FROM AUTHOR]

Precise targeting of transcription factor binding sites (TFBSs) is essential to comprehending transcriptional regulatory processes and investigating cellular function. Although several deep learning algorithms have been created to predict TFBSs, the models' intrinsic mechanisms and prediction results are difficult to explain. There is still room for improvement in prediction performance. We present DeepSTF, a unique deep-learning architecture for predicting TFBSs by integrating DNA sequence and shape profiles. We use the improved transformer encoder structure for the first time in the TFBSs prediction approach. DeepSTF extracts DNA higher-order sequence features using stacked convolutional neural networks (CNNs), whereas rich DNA shape profiles are extracted by combining improved transformer encoder structure and bidirectional long short-term memory (Bi-LSTM), and, finally, the derived higher-order sequence features and representative shape profiles are integrated into the channel dimension to achieve accurate TFBSs prediction. Experiments on 165 ENCODE chromatin immunoprecipitation sequencing (ChIP-seq) datasets show that DeepSTF considerably outperforms several state-of-the-art algorithms in predicting TFBSs, and we explain the usefulness of the transformer encoder structure and the combined strategy using sequence features and shape profiles in capturing multiple dependencies and learning essential features. In addition, this paper examines the significance of DNA shape features predicting TFBSs. The source code of DeepSTF is available at https://github.com/YuBinLab-QUST/DeepSTF/. [ABSTRACT FROM AUTHOR]

Gasification temperature measurement is one of the most challenging tasks in an entrained-flow gasifier and often requires indirect calculation using the soft-sensor method, a parameter prediction method using other parameters that are more easily measurable and using correlation equations that are widely accepted in the gasification field for the temperature data. Machine learning is a non-linear prediction method that can adequately act as a soft sensor. Furthermore, the recurrent neural network (RNN) has the function of memorization, which makes it capable of learning how to deal with temporal order. In this paper, the oxygen–coal ratio, CH4 content and CO2 content determined through the process analysis of a 3000-t/d coal-water slurry gasifier are used as input parameters for the soft sensor of the gasification temperature. The RNN model and back propagation (BP) neural network model are then established with training-set data from gasification results. Compared with prediction set data from the gasification results, the RNN model is found to be much better than the BP neural network based on important indexes such as the mean square error (MSE), mean absolute error (MAE) and standard deviation (SD). The results show that the MSE of the prediction set of the RNN model is 6.25°C, the MAE is 10.33°C and the SD is 3.88°C, respectively. The overall accuracy, the average accuracy and the stability effects are well within the accepted ranges for the results as such. [ABSTRACT FROM AUTHOR]

Even though numerous works focus on the few-shot learning issue by combining meta-learning, there are still limits to traditional graph classification problems. The antecedent algorithms directly extract features from the samples, and do not take into account the preference of the trained model to the previously "seen" targets. In order to overcome the aforementioned issues, an effective strategy with training an unbiased meta-learning algorithm was developed in this paper, which sorted out problems of target preference and few-shot under the meta-learning paradigm. First, the interactive attention extraction module as a supplement to feature extraction was employed, which improved the separability of feature vectors, reduced the preference of the model for a certain target, and remarkably improved the generalization ability of the model on the new task. Second, the graph neural network was used to fully mine the relationship between samples to constitute graph structures and complete image classification tasks at a node level, which greatly enhanced the accuracy of classification. A series of experimental studies were conducted to validate the proposed methodology, where the few-shot and semisupervised learning problem has been effectively solved. It also proved that our model has better accuracy than traditional classification methods on real-world datasets. [ABSTRACT FROM AUTHOR]

Objective Development of novel informatics methods focused on improving pregnancy outcomes remains an active area of research. The purpose of this study is to systematically review the ways that artificial intelligence (AI) and machine learning (ML), including deep learning (DL), methodologies can inform patient care during pregnancy and improve outcomes. Materials and methods We searched English articles on EMBASE, PubMed and SCOPUS. Search terms included ML, AI, pregnancy and informatics. We included research articles and book chapters, excluding conference papers, editorials and notes. Results We identified 127 distinct studies from our queries that were relevant to our topic and included in the review. We found that supervised learning methods were more popular (n  = 69) than unsupervised methods (n  = 9). Popular methods included support vector machines (n  = 30), artificial neural networks (n  = 22), regression analysis (n  = 17) and random forests (n  = 16). Methods such as DL are beginning to gain traction (n  = 13). Common areas within the pregnancy domain where AI and ML methods were used the most include prenatal care (e.g. fetal anomalies, placental functioning) (n  = 73); perinatal care, birth and delivery (n  = 20); and preterm birth (n  = 13). Efforts to translate AI into clinical care include clinical decision support systems (n  = 24) and mobile health applications (n  = 9). Conclusions Overall, we found that ML and AI methods are being employed to optimize pregnancy outcomes, including modern DL methods (n  = 13). Future research should focus on less-studied pregnancy domain areas, including postnatal and postpartum care (n  = 2). Also, more work on clinical adoption of AI methods and the ethical implications of such adoption is needed. [ABSTRACT FROM AUTHOR]

Protein arginine methylation is an important posttranslational modification (PTM) associated with protein functional diversity and pathological conditions including cancer. Identification of methylation binding sites facilitates a better understanding of the molecular function of proteins. Recent developments in the field of deep neural networks have led to a proliferation of deep learning-based methylation identification studies because of their fast and accurate prediction. In this paper, we propose DeepGpgs, an advanced deep learning model incorporating Gaussian prior and gated attention mechanism. We introduce a residual network channel to extract the evolutionary information of proteins. Then we combine the adaptive embedding with bidirectional long short-term memory networks to form a context-shared encoder layer. A gated multi-head attention mechanism is followed to obtain the global information about the sequence. A Gaussian prior is injected into the sequence to assist in predicting PTMs. We also propose a weighted joint loss function to alleviate the false negative problem. We empirically show that DeepGpgs improves Matthews correlation coefficient by 6.3% on the arginine methylation independent test set compared with the existing state-of-the-art methylation site prediction methods. Furthermore, DeepGpgs has good robustness in phosphorylation site prediction of SARS-CoV-2, which indicates that DeepGpgs has good transferability and the potential to be extended to other modification sites prediction. The open-source code and data of the DeepGpgs can be obtained from https://github.com/saizhou1/DeepGpgs. [ABSTRACT FROM AUTHOR]

In seismic interpretation, fault detection is a crucial step that often requires considerable manual labor and time. The convolutional neural network (CNN) is state-of-the-art deep learning technology that can perform even better than humans at image recognition. However, traditional methods of using CNNs for prediction require a very large dataset to train the network, which is impractical for common researchers and interpreters in geophysics who have difficulty obtaining sufficient quantities of labeled real data. In this paper, we propose a method for seismic fault detection using a CNN that requires only a very small training set. We treat the fault detection process as a semantic segmentation task and train an encoder–decoder CNN, namely, a U-Net, to perform a pixel-by-pixel prediction on the seismic section to determine whether each pixel is a fault or non-fault. Using this type of CNN in the experiments, we obtain good prediction results on real data. When interpreting a new seismic volume with the proposed method, interpreters need only to pick and label several 2D sections; subsequently, the model can predict faults in any other section of the same volume, greatly improving the interpretation efficiency. To evaluate the performance of the proposed method, we introduce a fault detection accuracy index that describes the accuracy of the prediction results. In this paper, we show that using only seven seismic sections of a seismic volume to train a CNN can allow us to predict faults successfully in any other section of the same volume. [ABSTRACT FROM AUTHOR]

We present a novel approach to identify galaxy clusters that are undergoing a merger using a deep learning approach. This paper uses massive galaxy clusters spanning 0 ≤ z ≤ 2 from The Three Hundred project, a suite of hydrodynamic resimulations of 324 large galaxy clusters. Mock, idealized Compton- y and X-ray maps were constructed for the sample, capturing them out to a radius of 2 R 200. The idealized nature of these maps mean they do not consider observational effects such as foreground or background astrophysical objects, any spatial resolution limits or restriction on X-ray energy bands. Half of the maps belong to a merging population as defined by a mass increase Δ M/M ≥ 0.75, and the other half serves as a controlled, relaxed population. We employ a convolutional neural network architecture and train the model to classify clusters into one of the groups. A best-performing model was able to correctly distinguish between the two populations with a balanced accuracy (BA) and recall of 0.77, ROC-AUC of 0.85, PR-AUC of 0.55, and F 1 score of 0.53. Using a multichannel model relative to a single-channel model, we obtain a 3 per cent improvement in BA score, and a 6 per cent improvement in F 1 score. We use a saliency interpretation approach to discern the regions most important to each classification decision. By analysing radially binned saliency values we find a preference to utilize regions out to larger distances for mergers with respect to non-mergers, greater than ∼1.2 R 200 and ∼0.7 R 200 for SZ and X-ray, respectively. [ABSTRACT FROM AUTHOR]

Deep neural networks are vulnerable to backdoor attacks. Previous backdoor attacks have mainly focused on images. Unlike images composed of regular pixels, 3D point clouds are composed of irregular three-dimensional XYZ coordinates, which are widely used in areas such as autonomous driving and 3D measurement. As many deep neural networks have been developed for processing 3D point clouds, these networks also face the risk of backdoor attacks. Nevertheless, backdoor attacks on 3D point clouds have rarely been investigated. This paper proposes a stealthy backdoor attack on point clouds in the physical world, aiming to generate trainable non-rigid deformations as backdoor patterns. Instead of directly adding backdoor patterns onto the point clouds, we deform the 3D space of the point clouds to a new space, ensuring that all point clouds have the same backdoor deformation. We use point cloud alignment to overcome the inconsistency of backdoor deformation caused by shifting and scaling in the physical world. We also propose a physical transformation layer to combat the physical transformations. Additionally, we propose mask contrast learning to eliminate pseudo backdoor patterns to make the network's backdoor property stealthier. Extensive experiments indicate that the proposed method can achieve better attack success rates and stealthiness. [ABSTRACT FROM AUTHOR]

Recent studies show that neural networks are vulnerable to backdoor attacks, in which compromised networks behave normally for clean inputs but make mistakes when a pre-defined trigger appears. Although prior studies have designed various invisible triggers to avoid causing visual anomalies, they cannot evade some trigger detectors. In this paper, we consider the stealthiness of backdoor attacks from input space and feature representation space. We propose a novel backdoor attack named sparse backdoor attack, and investigate the minimum required trigger to induce the well-trained networks to make incorrect results. A U-net-based generator is employed to create triggers for each clean image. Considering the stealthiness of the trigger, we restrict the elements of the trigger between −1 and 1. In the aspect of the feature representation domain, we adopt an entanglement cost function to minimize the gap between feature representations of benign and malicious inputs. The inseparability of benign and malicious feature representations contributes to the stealthiness of our attack against various model diagnosis-based defences. We validate the effectiveness and generalization of our method by conducting extensive experiments on multiple datasets and networks. [ABSTRACT FROM AUTHOR]

While tensor accelerated compilers have proven effective in deploying deep neural networks (DNN) on general-purpose hardware, optimizing for FPGA remains challenging due to the complex DNN architectures and the heterogeneous, semi-open compute units. This paper introduces the Automatic Kernel Generation for DNN on CPU-FPGA (AKGF) framework for efficient deployment of DNN on heterogeneous CPU-FPGA platforms. AKGF generates an intermediate representation (IR) of the DNN using TVM's Halide IR, annotates the operators of model layers in the IR to compute them on the corresponding hardware cores, and further optimizes the operator code for CPU and FPGA using ARM's function library and the polyhedral model to enhance model inference speed and power consumption. The experimental tests conducted on a CPU-FPGA board validate the effectiveness of AKGF, demonstrating significant acceleration ratios (up to 6.7x) compared to state-of-the-art accelerators while achieving a 2x power optimization. AKGF effectively leverages the computational capabilities of both CPU and FPGA for high-performance deployment of DNN on CPU-FPGA platforms. [ABSTRACT FROM AUTHOR]

Knowledge graph (KG) is introduced as side information into recommender systems, which can alleviate the sparsity and cold start problems in collaborative filtering. Existing studies mainly focus on modeling users' historical behavior data and KG-based propagation. However, they have the limitation of ignoring noise information during recommendation. We consider that noise exists in two parts (i.e. KG and user-item interaction data). In this paper, we propose Knowledge-aware Dual-Channel Graph Neural Networks (KDGNN) to improve the recommendation performance by reducing the noise in the recommendation process. Specifically, (1) for the noise in KG, we design a personalized gating mechanism, namely dual-channel balancing mechanism, to block the propagation of redundant information in KG. (2) For the noise in user-item interaction data, we integrate personalized and knowledge-aware signals to capture user preferences fully and use personalized knowledge-aware attention to denoise user-item interaction data. Compared with existing KG-based methods, we aim to propose a knowledge-aware recommendation method from a new perspective of denoising. We perform performance analysis on three real-world datasets, and experiment results demonstrate that KDGNN achieves strongly competitive performance compared with several compelling state-of-the-art baselines. [ABSTRACT FROM AUTHOR]

This paper considers the application of artificial neural network (ANN) models for electric field intensity and magnetic flux density estimation in the proximity of overhead transmission lines. Specifically, two distinct ANN models are used to facilitate independent estimation of electric field intensity and magnetic flux density in the proximity of overhead transmission lines. The considered ANN approach is systematically evaluated under different scenarios. An example of an overhead transmission line with horizontal phase conductor configuration is used to enable a direct comparison of the electric field intensity and magnetic flux density estimates generated by the two ANN models to measurement results obtained over the lateral profile. Further investigation of ANN models involves an extensive study whereby 13 different overhead transmission lines of horizontal configurations are used as the basis for comparing measurement results to estimates provided by the ANN models. In this study, the performance analysis of the ANN models was evaluated using coefficient of determination and root mean square error. The obtained results demonstrate that the considered ANN approach can be used to estimate the electric field intensity and magnetic flux density in the proximity of overhead transmission lines. [ABSTRACT FROM AUTHOR]

The inference of gene regulatory networks (GRNs) is of great importance for understanding the complex regulatory mechanisms within cells. The emergence of single-cell RNA-sequencing (scRNA-seq) technologies enables the measure of gene expression levels for individual cells, which promotes the reconstruction of GRNs at single-cell resolution. However, existing network inference methods are mainly designed for data collected from a single data source, which ignores the information provided by multiple related data sources. In this paper, we propose a multi-view contrastive learning (DeepMCL) model to infer GRNs from scRNA-seq data collected from multiple data sources or time points. We first represent each gene pair as a set of histogram images, and then introduce a deep Siamese convolutional neural network with contrastive loss to learn the low-dimensional embedding for each gene pair. Moreover, an attention mechanism is introduced to integrate the embeddings extracted from different data sources and different neighbor gene pairs. Experimental results on synthetic and real-world datasets validate the effectiveness of our contrastive learning and attention mechanisms, demonstrating the effectiveness of our model in integrating multiple data sources for GRN inference. [ABSTRACT FROM AUTHOR]

Recurrent neural networks based on reservoir computing are increasingly being used in many applications. Optimization of the topological structure of the reservoir and the internal connection weights for a given task is one of the most important problems in reservoir computing. In this paper, considering the fact that one can construct a large matrix using Kronecker products of several small-size matrices, we propose a method to optimize the reservoir. Having a small number of parameters to optimize, a gradient based algorithm is applied to optimize parameters, and consequently the reservoir. In addition to reducing the number of parameters for optimization, potentially, the method is able to control several other properties of the reservoir such as spectral radius, sparsity, weight distribution and underlying connections, i.e. connection topology. To reveal the effectiveness of the proposed optimization method, the application to the following tasks are considered: Nonlinear autoregressive moving average and multiple superimposed oscillators. Simulation results show satisfactory performance of the method. [ABSTRACT FROM PUBLISHER]

In the Internet of Things, smart devices are expected to correctly capture and process data from environments, regardless of perturbation and adversarial attacks. Therefore, it is important to guarantee the robustness of their intelligent components, e.g. neural networks, to protect the system from environment perturbation and adversarial attacks. In this paper, we propose a formal verification technique for rigorously proving the robustness of neural networks. Our approach leverages a tight liner approximation technique and constraint substitution, by which we transform the robustness verification problem into an efficiently solvable linear programming problem. Unlike existing approaches, our approach can automatically generate adversarial examples when a neural network fails to verify. Besides, it is general and applicable to more complex neural network architectures such as CNN, LeNet and ResNet. We implement the approach in a prototype tool called WiNR and evaluate it on extensive benchmarks, including Fashion MNIST, CIFAR10 and GTSRB. Experimental results show that WiNR can verify neural networks that contain over 10 000 neurons on one input image in a minute with a 6.28% probability of false positive on average. [ABSTRACT FROM AUTHOR]

Collaborative filtering (CF) is a well-known and eminent recommendation technique to predict the preference of new users by revealing the structures of historical records of the examined users. Even though CF is effectively adapted in several commercial areas, many limitations still exist, particularly in the sparsity of rating data that raises many issues. This paper devises a novel deep learning strategy for CF to recognize user preferences. Here, black hole entropic fuzzy clustering (BHEFC) is devised for clustering item sequences to form groups with similar item sequences. Moreover, cluster centroids are optimized using the tunicate swarm magnetic optimization algorithm (TSMOA), which is devised by combining tunicate swarm algorithm and magnetic optimization algorithm. After grouping similar items together, the group matching is performed based on a deep convolutional neural network (Deep CNN). Subsequently, the visitor sequence and query sequence are compared using Jaro–Winkler distance, which contributes to the best visitor sequence. From this best visitor sequence, the recommended product is acquired. The proposed TSMOA–BHEFC and Deep CNN outperformed other methods with minimal mean absolute error of 0.200, mean absolute percentage error of 0.198 and root mean square error of 0.447, respectively. [ABSTRACT FROM AUTHOR]

Today, machine learning plays a major role in different branches of the healthcare industry, from prognosis and diagnosis to drug development providing a significant perspective on the medical landscape for disease prevention or treatment and the improvement of human life. Recently, the use of deep neural networks in different machine learning applications has shown a great contribution to the improvement of the accuracy of predictions. In this paper, a novel application of convolutional neural networks on medical prognosis is presented. The proposed method employs a one-dimensional convolutional neural network (1D-CNN) to predict the survivability of breast cancer patients. After further examining the network architecture, a number of 8, 14 and 24 convolutional filters were considered within three layers, respectively, followed by a max-pooling layer after the second and third layers. In addition, regarding the probabilistic nature of the survivability prediction problem, an extra layer was added to the network in order to calculate the probability of the patient survivability. To train the developed 1D-CNN machine, the SEER database as the most reliable repository of cancer survivability was used to retrieve the required training set. After a pre-processing to remove unusable records, a set of 50 000 breast cancer cases including 35 features was prepared for training the machine. Based on the results obtained in this study, the developed machine could reach an accuracy of 85.84%. This accuracy is the highest level of accuracy compared to the previous prediction methods. Furthermore, the mean squared error of the calculated probability was 0.112, which is an acceptable value of error for a probability calculation machine. The output of the developed machine can be used reliably by physicians to make decision about the most appropriate treatment strategy. [ABSTRACT FROM AUTHOR]

The advent of fast sensing technologies allow for real-time model updates in many applications where the model parameters are uncertain. Once the observations are collected, Bayesian algorithms offer a pathway for real-time inversion (a.k.a. model parameters/inputs update) because of the flexibility of the Bayesian framework against non-uniqueness and uncertainties. However, Bayesian algorithms rely on the repeated evaluation of the computational models and deep learning (DL) based proxies can be useful to address this computational bottleneck. In this paper, we study the effects of the approximate nature of the deep learned models and associated model errors during the inversion of borehole electromagnetic (EM) measurements, which are usually obtained from logging while drilling. We rely on the iterative ensemble smoothers as an effective algorithm for real-time inversion due to its parallel nature and relatively low computational cost. The real-time inversion of EM measurements is used to determine the subsurface geology and properties, which are critical for real-time adjustments of the well trajectory (geosteering). The use of deep neural network (DNN) as a forward model allows us to perform thousands of model evaluations within seconds, which is very useful to quantify uncertainties and non-uniqueness in real-time. While significant efforts are usually made to ensure the accuracy of the DL models, it is widely known that the DNNs can contain some type of model-error in the regions not covered by the training data, which are unknown and training specific. When the DL models are utilized during inversion of EM measurements, the effects of the model-errors could manifest themselves as a bias in the estimated input parameters and as a consequence might result in a low-quality geosteering decision. We present numerical results highlighting the challenges associated with the inversion of EM measurements while neglecting model-error. We further demonstrate the utility of a recently proposed flexible iterative ensemble smoother in reducing the effect of model-bias by capturing the unknown model-errors, thus improving the quality of the estimated subsurface properties for geosteering operation. Moreover, we describe a procedure for identifying inversion multimodality and propose possible solutions to alleviate it in real-time. [ABSTRACT FROM AUTHOR]

Facial expression recognition (FER) is the process of identifying human expressions. People vary in their accuracy at recognizing the emotions of others. Use of technology to help people with emotion recognition is a relatively important research area. Various works have been conducted on automating the recognition of facial expressions. The main intent of this paper is to plan for the FER model with the aid of intelligent techniques. The proposed models consist of steps like data collection, face detection, optimized feature extraction and emotion recognition. Initially, the standard benchmark facial emotion dataset is collected, and it is subjected to face detection. The optimized scale-invariant feature transform (OSIFT) is adopted for feature extraction, in which the key points that are giving unique information are optimized by the hybrid meta-heuristic algorithm. Two meta-heuristic algorithms like spotted hyena optimization and beetle swarm optimization (BSO) are merged to form the proposed spotted hyena-based BSO (SH-BSO). Also, the local tri-directional pattern is extracted, which is further combined with optimized SIFT. Here, the proposed SH-BSO is utilized for optimizing the number of hidden neurons of both deep neural network and convolutional neural network in such a way that the recognition accuracy could attain maximum. [ABSTRACT FROM AUTHOR]

A large amount of clinical evidence began to mount, showing that circular ribonucleic acids (RNAs; circRNAs) perform a very important function in complex diseases by participating in transcription and translation regulation of microRNA (miRNA) target genes. However, with strict high-throughput techniques based on traditional biological experiments and the conditions and environment, the association between circRNA and miRNA can be discovered to be labor-intensive, expensive, time-consuming, and inefficient. In this paper, we proposed a novel computational model based on Word2vec, Structural Deep Network Embedding (SDNE), Convolutional Neural Network and Deep Neural Network, which predicts the potential circRNA-miRNA associations, called Word2vec, SDNE, Convolutional Neural Network and Deep Neural Network (WSCD). Specifically, the WSCD model extracts attribute feature and behaviour feature by word embedding and graph embedding algorithm, respectively, and ultimately feed them into a feature fusion model constructed by combining Convolutional Neural Network and Deep Neural Network to deduce potential circRNA-miRNA interactions. The proposed method is proved on dataset and obtained a prediction accuracy and an area under the receiver operating characteristic curve of 81.61% and 0.8898, respectively, which is shown to have much higher accuracy than the state-of-the-art models and classifier models in prediction. In addition, 23 miRNA-related circular RNAs (circRNAs) from the top 30 were confirmed in relevant experiences. In these works, all results represent that WSCD would be a helpful supplementary reliable method for predicting potential miRNA-circRNA associations compared to wet laboratory experiments. [ABSTRACT FROM AUTHOR]