Automatic Search Research Articles

Mosaic: An Interoperable Compiler for Tensor Algebra

We introduce Mosaic, a sparse tensor algebra compiler that can bind tensor expressions to external functions of other tensor algebra libraries and compilers. Users can extend Mosaic by adding new functions and bind a sub-expression to a function using a scheduling API. Mosaic substitutes the bound sub-expressions with calls to the external functions and automatically generates the remaining code using a default code generator. As the generated code is fused by default, users can productively leverage both fusion and calls to specialized functions within the same compiler. We demonstrate the benefits of our dual approach by showing that calling hand-written CPU and specialized hardware functions can provide speedups of up to 206× against fused code in some cases, while generating fused code can provide speedups of up to 3.57× against code that calls external functions in other cases. Mosaic also offers a search system that can automatically map an expression to a set of registered external functions. Both the explicit binding and automatic search are verified by Mosaic. Additionally, the interface for adding new external functions is simple and general. Currently, 38 external functions have been added to Mosaic, with each addition averaging 20 lines of code.

Analysis of possible reasons of erronious identificationof the dynamic parameters of a controlled thermal plant

For optimal adjustment of the temperature regulator, it is necessary to know the dynamic properties of the controlled plant, which are mainly obtained from an experimen, in the procedure of identifying the dynamic parameters of the controlled plant. The purpose of the study is to analyze possible causes of erronious identification of dynamic parameters of the plant. A possible mistake may be an unsuccessful selection of parameters, being identified. Their necessary and sufficient list includes static coefficient of transmission, the main and additional time constants, time delay. Possible strong nonlinearity of the actuator should be taken into account to compensate it in calculations. During the experiment, it is desirable not to allow large temperature deviations from the normal operating mode, to avoid the influence of possible small nonlinearity of the controlled plant. To process the results of the experiment, verified universal automatic optimization algorithms should be used. As a minimization criterion, it is recommended to use the logarithm of the mean square deviation between the experimental temperature oscillogram and the corresponding calculated oscillogram on the model. If, as the result of identification, equal or close values of the main time constants are obtained, this may be the result of a fault of the automatic search procedure, which is "stuck" in a local minimax. The values should be divercified, to help the automatic optimizer.

A real-time mechanical fault diagnosis approach based on lightweight architecture search considering industrial edge deployments

Mechanical intelligence diagnostic models based on deep learning have increased diagnostic accuracy. However, the industrial application of deep-learning models is hindered by the extra memory requirement and the increasing computational overhead of deep-learning models, which limit edge deployment and reduce real-time performance, respectively. Therefore, we propose a mechanical fault diagnosis strategy based on a multi-objective automatic optimization architecture search that combines variable layer search, efficient search space, and a real-time search strategy. A lightweight diagnostic network that integrates the computing requirements of edge deployment, real-time testing, and accuracy is constructed. The proposed model considers the influence of compound faults and the strong noise of rotating machinery in engineering applications. The experiments on the defect data of two sets of rotating equipment show that our model is more lightweight and has higher accuracy than other advanced baselines.

A multi-objective memetic algorithm for automatic adversarial attack optimization design

The phenomenon of adversarial examples has been revealed in variant scenarios. Recent studies show that well-designed adversarial defense strategies can improve the robustness of deep learning models against adversarial examples. However, with the rapid development of defense technologies, it also tends to be more difficult to evaluate the robustness of the defensed model due to the weak performance of existing manually designed adversarial attacks. To address the challenge, given the defensed model, the efficient adversarial attack with less computational burden and lower robust accuracy is needed to be further exploited. Therefore, we propose a multi-objective memetic algorithm for auto adversarial attack optimization design, which realizes the automatic search for the near-optimal adversarial attack towards defensed models. Firstly, the more general mathematical model of auto adversarial attack optimization design is constructed, where the search space includes not only the attacker operations, magnitude, iteration number, and loss functions but also the connection ways of multiple adversarial attacks. In addition, we develop a multi-objective memetic algorithm combining NSGA-II and local search to solve the optimization problem. Finally, to decrease the evaluation cost during the search, we propose a representative data selection strategy based on sorting cross entropy loss values of images output by models. Experiments on CIFAR10, CIFAR100, and ImageNet datasets show that our method can efficiently find near-optimal adversarial attacks with lower robust accuracy and less time cost, which can provide more reliable and efficient robustness evaluation for defensed models.

SAT-Inspired Higher-Order Eliminations

We generalize several propositional preprocessing techniques to higher-order logic, building on existing first-order generalizations. These techniques eliminate literals, clauses, or predicate symbols from the problem, with the aim of making it more amenable to automatic proof search. We also introduce a new technique, which we call quasipure literal elimination, that strictly subsumes pure literal elimination. The new techniques are implemented in the Zipperposition theorem prover. Our evaluation shows that they sometimes help prove problems originating from Isabelle formalizations and the TPTP library.

Automatic tracking method for submarine cables and pipelines of AUV based on side scan sonar

Submarine cables and pipelines laid on the seafloor are critical infrastructures for energy transportation and communication transmission and must be inspected in a timely manner to determine maintenance needs. Autonomous inspection of cables (including pipelines) and their surroundings using underwater vehicles is difficult because of sensor myopia and deviations in cable routing. This study proposes an automatic cable-tracking method (ACTM) based on side-scan sonar (SSS) that endows autonomous underwater vehicles (AUVs) with autonomous decision-making capabilities to achieve fast localization and stable tracking of submarine cables with large uncertainties. Compared with the traditional predefined waypoint method, the proposed method decomposes the tasks into a series of independent behaviors according to operational objectives; thus, the AUV can perform online dynamic task replanning and realize an autonomous coverage search of the target region. When a cable is detected, the cable-tracking task is constructed as a path-following control problem on the horizontal plane. To reduce uncertainties caused by environmental noise and sensor measurement errors, a uniform motion model based on target tracking theory was used to model the virtual mass abstracted from the cable, and the states of the cable were estimated using the Kalman filter. Subsequently, an adaptive line-of-sight guidance algorithm for the SSS is established to dynamically adjust the look-ahead distance by introducing states such as cable curvature and angular change rate, which somewhat alleviates the tracking loss caused by the SSS short-sightedness. The proposed ACTM effectiveness and robustness were validated through numerical simulations and field tests. The results showed that the proposed method can achieve automatic search and adaptive tracking of cables, further reducing the role of humans in the task.

ForeTiS: A comprehensive time series forecasting framework in Python

Summary: Time series forecasting is a research area with applications in various domains, nevertheless without yielding a predominant method so far. We present ForeTiS, a comprehensive and open source Python framework that allows rigorous training, comparison, and analysis of state-of-the-art time series forecasting approaches. Our framework includes fully automated yet configurable data preprocessing and feature engineering. In addition, we use advanced Bayesian optimization for automatic hyperparameter search. ForeTiS is easy to use, even for non-programmers, requiring only a single line of code to apply state-of-the-art time series forecasting. Various prediction models, ranging from classical forecasting approaches to machine learning techniques and deep learning architectures, are already integrated. More importantly, as a key benefit for researchers aiming to develop new forecasting models, ForeTiS is designed to allow for rapid integration and fair benchmarking in a reliable framework. Thus, we provide a powerful framework for both end users and forecasting experts.Availability:ForeTiS is available at https://github.com/grimmlab/ForeTiS. We provide a setup using Docker, as well as a Python package at https://pypi.org/project/ForeTiS/. Extensive online documentation with hands-on tutorials and videos can be found at https://foretis.readthedocs.io.

Verification-Preserving Inlining in Automatic Separation Logic Verifiers

Bounded verification has proved useful to detect bugs and to increase confidence in the correctness of a program. In contrast to unbounded verification, reasoning about calls via (bounded) inlining and about loops via (bounded) unrolling does not require method specifications and loop invariants and, therefore, reduces the annotation overhead to the bare minimum, namely specifications of the properties to be verified. For verifiers based on traditional program logics, verification is preserved by inlining (and unrolling): successful unbounded verification of a program w.r.t. some annotation implies successful verification of the inlined program. That is, any error detected in the inlined program reveals a true error in the original program. However, this essential property might not hold for automatic separation logic verifiers such as Caper, GRASShopper, RefinedC, Steel, VeriFast, and verifiers based on Viper. In this setting, inlining generally changes the resources owned by method executions, which may affect automatic proof search algorithms and introduce spurious errors. In this paper, we present the first technique for verification-preserving inlining in automatic separation logic verifiers. We identify a semantic condition on programs and prove in Isabelle/HOL that it ensures verification-preserving inlining for state-of-the-art automatic separation logic verifiers. We also prove a dual result: successful verification of the inlined program ensures that there are method and loop annotations that enable the verification of the original program for bounded executions. To check our semantic condition automatically, we present two approximations that can be checked syntactically and with a program verifier, respectively. We implement these checks in Viper and demonstrate that they are effective for non-trivial examples from different verifiers.

Meet‐in‐the‐middle attacks on round‐reduced CRAFT based on automatic search

AbstractCRAFT is a lightweight block cipher designed by Beierle et al. to effectively resist differential fault attacks at fast software encryption 2019. In this article, Demirci‐Selçuk meet‐in‐the‐middle (DS‐MITM) attacks on round‐reduced CRAFT based on automatic search are proposed. A DS‐MITM automatic search model for CRAFT was constructed, and then, the automatic search model was used to detect a 9‐round DS‐MITM distinguisher. The strong relations between the round‐subtweakeys were observed and the key‐dependent sieve technique was adopted to reduce the memory complexity of the attack. Based on the 9‐round distinguisher, a 19‐round DS‐MITM attack can be presented. Due to the strong key relations, the time complexity can be reduced by the key‐bridging technique and the equivalent round‐subtweakey. The time complexity of the 19‐round DS‐MITM attack is 2114.68 19‐round CRAFT encryption, the data complexity is 256 chosen plaintexts, and the memory complexity is 2109 64‐bit blocks. Adding one round to the end of the 19‐round DS‐MITM attack, a 20‐round DS‐MITM attack can be proposed. The time complexity of the 20‐round attack is 2126.94 20‐round CRAFT encryption, the data complexity is 256 chosen plaintexts, and the memory complexity is 2109 64‐bit blocks.

EEG-based Emotion Recognition via Transformer Neural Architecture Search

Emotion recognition based on electroencephalogram (EEG) plays an increasingly important role in the field of brain–computer interfaces. Recently, deep learning has been widely applied to EEG decoding owning to its excellent capabilities in automatic feature extraction. Transformer holds great superiority in processing time-series signals due to its long-term dependencies extraction ability. However, most existing transformer architectures are designed manually by human experts, which is a time-consuming and resource-intensive process. In this article, we propose an automatic transformer neural architectures search (TNAS) framework based on multiobjective evolution algorithm (MOEA) for the EEG-based emotion recognition. The proposed TNAS conducts the MOEA strategy that considers both accuracy and model size to discover the optimal model from well-trained supernet for the emotion recognition. We conducted extensive experiments to evaluate the performance of the proposed TNAS on the DEAP and DREAMER datasets. The experimental results showed that the proposed TNAS outperforms the state-of-the-art methods.

Information technologies for automation of hyddrophisical situation molelling and forecasting in the Black Sea region

The article offers a solution to the scientific and applied problem of automating the acquisition and analysis of a certain set of hydrophysical indicators with the subsequent formation of a forecast of the hydrophysical situation in the Black Sea water area. This topic is of great relevance both in peacetime and in the conditions of the current armed aggression of the Russian Federation against Ukraine. The solved task is an important element of solving the more complex problem of operational obtaining of oceanographic data for highlighting the hydrographic situation in the Black Sea water area, collection, accumulation and visualization of oceanographic data in the interests of navigation and hydrographic support of users of such information. To solve this problem, a software environment was created for automatic search and selection of hydrophysical data, as well as their automatic input into specialized software for modeling and forecasting the hydrophysical situation with subsequent calculation of forecasts of hydrophysical fields. A set of the most important data was formed, which includes indicators of salinity, temperature, circulation. Automatic search and selection of meteorological and hydrophysical data takes place from international GFS and COPERNICUS profile databases. Also, to increase the effect of the simulations and forecasts, a specialized interface was developed for searching, visualization, and the possibility of downloading by the user the results of the forecast module of the mathematical modeling of the spatio-temporal variability of the hydrophysical fields of the northwestern part of the Black Sea. The developed scientific and practical approaches and toolkit are designed to function within the framework of the distributed and centralized Bank of Oceanographic Data of the National Academy of Sciences of Ukraine, part of which is an intellectual information system for highlighting the hydrographic situation in the Black Sea water area. The obtained results are also aimed at increasing the level of environmental safety in the Black Sea water area and can be used to monitor nature management in this region.

Prediction of maximum tensile stress in plain-weave composite laminates with interacting holes via stacked machine learning algorithms: A comparative study

Plain weave composite is a long-lasting type of fabric composite that is stable enough when being handled. Open-hole composites have been widely used in industry, though they have weak structural performance and complex design processes. An extensive number of material/geometry parameters have been utilized for designing these composites, thereby an efficient computational tool is essential for that purpose. Different Machine Learning (ML) tools were integrated to obtain the model with the highest accuracy considering the maximum tensile stress of composite plates with two interacting notches while comparing the effectiveness of each technique. Finite Element (FE) simulations were carried out inside the ABAQUS software by employing python macro code to provide a data-rich framework (8960 data). The predictions given by ML methods were compared with the data given by the numerical simulations. An evolutionary algorithm (TPOT) and automatic neural network search (AuoKeras) were utilized for that purpose. An automatic grid search technique was employed to select the best method which could predict the material attribute target values (maximum stress) for different tests. 1% of the data was given as training while 99% was for testing to ensure the robustness of the model. It was concluded that the model containing the Gradient Boosting Regression (GBR), PolyFeatures, and LassoLarsCV algorithms outperformed other ML combinations and Artificial Neural Networks (ANN) for predicting the target value. The coefficient of determination (R2) and root mean square error (RMSE) of the proposed model were 0.97 and 253 respectively. Hence, this model could be utilized for prospective predictions in this type of materials and geometry by providing further reduction of the computational time and labor cost with high accuracy.

NEURAL NETWORK SUPPORT FOR AUTOMATION OF OBJECT CONTROL WITH MULTIFACTORIAL INPUT UNDER CONDITIONS OF UNCERTAINTY

High dimension of the factors, their noisiness, subjectivity of the human factor breed high uncertainty in the implementation of effective monitoring and productive management in the areas of production and consumption, and hinder the optimal decisions making. In these conditions, it is appropriate to apply intelligent data analysis procedures based on artificial neural networks. Purpose of the study is to substantiate the technology for constructing of effective neural network models for automatic assessment of the object states and their control by finding the optimal values of input factors based on the analysis of the initial set of retrospective data. The desired product is set of neural network models for simultaneous assessment of the current object states and the calculation of the input factors values that ensure the achievement of the required objective function indicators. To automate the processes of recognizing the states of the study object and adapting the factors that bring the current state to the target one, a functional dependence of the states and factors is found based on forced learning of a synthesized models ensemble. The proposed technology and technical tools make it possible to automate the processes of classifying the states of the study objects, adapt the input factors to the target states, and evaluate the quality by model testing. The practical significance of the study results is in the creation of a universal toolkit for a whole class of objects in the tasks of automatic state classification and search for input factors space that is adequate to the target state space. Functionally, the ensemble of trained models can be implemented as a data analysis software unit in the format of two subsystems: for recognizing the states of the study object and adapting input factors to target states. Automation of the state classifying tasks and adapting the input set in the proposed technology that is performed on the basis of standard technical data analysis packages, makes it possible to increase the efficiency of decision-making and reduce financial costs in the implementation of industrial and commercial projects.

The role of assistive technology in supporting the engagement of students with visual impairment in learning mathematics: An integrative literature review

This integrative literature review examined how assistive technology (AT) can help students with visual impairment (VI) in middle primary and secondary schools (aged 11–18 years) engage in learning Mathematics. To collate the relevant findings across studies, an integrative literature review approach was employed and supported by an evidence-based synthesis method. Automatic search phrases were used to find research articles related to the topic in the Google Scholar, Web of Science, Scopus, and Springer Link databases. A total of 628 articles were identified during the initial search of which only 11 were included after putting them under rigorous selection criteria. In addition, a mixed-method appraisal technique (MMAT) was used to examine the methodological quality of the papers that were included. The synthesis result came up with three overarching themes: the availability of AT to support learning mathematics, the usability of AT in mathematics classes, and the role of AT in the inclusion of visually impaired students. The findings based on the results of the 11 academic articles published from 2007 to 2021 revealed that effective engagement of students with VIs in Mathematics instruction is achieved by providing students with VI equitable access to the swiftly evolving AT, addressing issues related to affordability, testing the usability and appropriateness of existing AT and enhancing accessibility of AT for all by breaking the social stigma against students with VI in using AT. It is concluded that easily affordable and accessible assistive technologies with multiple non-visual display formats that enable students with VI to access mathematical symbols, notations, expressions, and tactual contents should be in place to actively engage students with VI in learning mathematics.

SAT-aided Automatic Search of Boomerang Distinguishers for ARX Ciphers

In Addition-Rotation-Xor (ARX) ciphers, the large domain size obstructs the application of the boomerang connectivity table. In this paper, we explore the problem of computing this table for a modular addition and the automatic search of boomerang characteristics for ARX ciphers. We provide dynamic programming algorithms to efficiently compute this table and its variants. These algorithms are the most efficient up to now. For the boomerang connectivity table, the execution time is 42(n − 1) simple operations while the previous algorithm costs 82(n − 1) simple operations, which generates a smaller model in the searching phase. After rewriting these algorithms with boolean expressions, we construct the corresponding Boolean Satisfiability Problem models. Two automatic search frameworks are also proposed based on these models. This is the first time bringing the SAT-aided automatic search techniques into finding boomerang attacks on ARX ciphers. Finally, under these frameworks, we find out the first verifiable 10-round boomerang trail for SPECK32/64 with probability 2−29.15 and a 12-round trail for SPECK48/72 with probability 2−44.15. These are the best distinguishers for them so far. We also perceive that the previous boomerang attacks on LEA are constructed with an incorrect computation of the boomerang connection probability. The result is then fixed by our frameworks.

Software Code Refactoring: A Comprehensive Review

The complexity of software has increased because of the development as well as the difficulty of requirements during the development of software, or to add new features that eventually lead to reduce the quality of the software as a whole. Software refactoring can be defined as included processes in the maintenance period of a software life cycle, and it is a technique to clean the software code from code bad smell and to improve the internal structure of the software, in addition to increasing the quality of software by using a set of activities without changing the external behavior of a software. Researchers have been developing techniques to reform software during the code or design standard to decrease the effort and time required for maintenance processes. This paper provides a systematic review of the literature for 17 studies of code standards, An automatic search is utilized in the digital libraries to look for the relevant studies that were published from 2014 to 2021, the best five researchers are chosen in this subject, five studies or less are chosen for each depending on the number of reference in the database of scientific sites, or using an approach or a new method to get good results. Eventually, each paragraph is analyzed and mentions the method or algorithm used in rebuilding software, further the aims, and the result for each paper.

AutoMSNet: Multi-Source Spatio-Temporal Network via Automatic Neural Architecture Search for Traffic Flow Prediction

Recently the research of traffic flow prediction with deep learning framework has be largely developed, whereas most current methods are still faced with the following shortcomings. For spatial feature extraction, studies have shown that both local and non-local correlations exist on traffic networks. Considering the temporal dependencies, short-term impending and longer periodic components are two most critical patterns of traffic data, which further provide different information for the prediction task. Furthermore, multi-source heterogeneous external data, which naturally holds semantic gap with traffic data, also have impact on traffic flow. To solve the above problems, this paper proposes an AutoMSNet (Multi-Source Spatio-Temporal Network via Automatic neural architecture search). The AutoMSNet is composed of an encoder-decoder structure. The encoder takes neighboring data as inputs, while the decoder captures long-term periodic patterns. Thus, different functions of two temporal features are simultaneously extracted. Moreover, a neural architecture search space is designed for spatial feature extraction. Through architecture search technique, graph convolutions with different receptive fields are automatically selected and combined to form an optimal module structure. Therefore, both local and non-local spatial features can be adaptively captured. Besides, a meta learning feature fusion strategy is proposed to integrate external data, which can alleviate the semantic gap between different data sources. Extensive experiments on three real-world traffic datasets evaluate the superiority of the proposed model.

Rapid Location of Remote Volcanic Infrasound Using 3D Ray Tracing and Empirical Climatologies: Application to the 2011 Cordón Caulle and 2015 Calbuco Eruptions, Chile

AbstractInfrasound (<20 Hz) can propagate thousands of kilometers through the atmosphere, enabling global source location using networks of arrays such as the International Monitoring System infrasound network. However, atmospheric spatiotemporal variability poses a major challenge to locating infrasound sources. Strong horizontal cross‐winds deviate the observed infrasound arrival azimuths, producing source mislocations that can be on the order of hundreds of kilometers. We introduce a method that combines empirical climatologies (HWM14/NRLMSIS2.0) and 3D ray tracing (infraGA) to obtain first‐order, robust, and rapid estimates of the backazimuth deviations for source location procedures. For each trial source node and receiver path, day of the year and time, we perform an automatic iterative search for infrasound ground intercepts around the target station and obtain the corresponding backazimuth deviation. We test the method using IMS infrasound data from stations up to ∼5,000 km range for two similar explosive eruption case studies: 2011 Puyehue‐Cordón Caulle and 2015 Calbuco. We obtain a source mislocation reduction up to ∼84% (242–38.7 km) and ∼75% (366–93.1 km) for Puyehue‐Cordón Caulle and Calbuco eruptions, respectively. To evaluate the approach, we repeat the procedure using more realistic hybrid atmospheric descriptions; we obtain comparable results (up to ∼75% mislocation reduction for both eruptions). Potential applications of the approach include long‐range volcano monitoring in near‐real time by using pre‐computed look‐up tables or large‐scale, multi‐year reanalyses of infrasound waveform archives.

Digital technologies for mental health improvements in the COVID-19 pandemic: a scoping review

Digital technologies have been used to support mental health services for two decades, but the COVID-19 pandemic created a particular opportunity for greater utilization and more data-driven assessment of these digital technologies. This research aims to offer a scoping review of the characteristics and effectiveness of digital interventions that were employed to improve mental health in the real context of COVID-19 pandemic. A combination of search terms was applied for automatic search of publications in the relevant databases. The key features of included studies were extracted, including the intervention, participant, and study details. A total of 20 eligible studies were included in the final review, which were conducted across different geographic regions and among diverse cultural groups. Among them, fourteen studies mainly reported the impact of digital technologies on general population, while only one published study developed specific interventions for the isolated COVID-19 depressed patients in hospitals. Digital technologies identified in this review were mainly developed via web-based and mobile-based platforms, such as social networking and video conferencing applications. But less than half of them were aligned with theoretical approaches from standardized psychological treatments. Most of the studies have reported positive effects of digital technologies, either on improving general mental and emotional well-being or addressing specific conditions (e.g., depression, stress, and anxiety). This scoping review suggests that digital technologies hold promise in bridging the mental health-care gap during and after the COVID-19 pandemic, and calls for more rigorous studies to identify pertinent features that are likely to achieve more effective mental health outcomes.

Joint spatio-temporal features and sea background prior for infrared dim and small target detection

The detection of dim and small targets in the complex sea background poses new challenges to automatic target search and interception by infrared air-to-air missiles at sea. To solve the problems of low detection rate of dim targets and massive suspected targets caused by the shadow region of clutter on the sea surface, this paper proposes a detection method for dim and small targets with spatio-temporal features and background prior information. Firstly, by combining the ideas of local energy factor and multi-scale patch contrast measure, an improved dim target detection method is proposed, which can effectively enhance the contrast of dim targets and suppress background clutter. Then, according to the grayscale probability distribution difference between the background and the target, a double-Gaussian model is established for the residual background and the target region and a background suppression method based on the spatial prior information is proposed to further suppress the residual background. Finally, a dim target detection method is proposed by combining the above dim target detection, background suppression, and dynamic pipeline filtering methods. The results indicate the average detection rate of the detection method proposed in this study is 91.44%. Also, the average number of false alarms in a single frame is reduced to 9.89, which is more than 80% lower than that of other algorithms.