Original Source Code Research Articles

Analysis and Experimentation on the ManTraNet Image Forgery Detector

This work describes the ManTraNet network for image forgery detection. ManTraNet is an end-to-end convolutional neural network composed of two sub-networks, one to extract features linked to traces of manipulation, and another to detect local anomalies between the features. It is trained on pristine and forged images from several datasets. We briefly analyze the results provided by ManTraNet, so as to highlight its qualities and limitations. Overall, ManTraNet yields state-of-the-art results on benchmark datasets with images similar to the one it sees in training, but is unreliable on wild images, due to its opacity and the difficulty distinguishing true detections from false positives. **This is an MLBriefs article, the source code has not been reviewed!**<br> **The original source code is [[available here|https://github.com/RonyAbecidan/ManTraNet-pytorch]] (last checked 2022/10/19).**

Experiments on Deep Single-Image Portrait Relighting

This work describes deep single-image portrait relighting, a method to change the lighting of an image. The method is based on an hourglass convolutional neural network, which encodes the image and its estimated original lighting features, then replaces the lighting features with the target light. We highlight the overall good results produced by this method, as well as its limitations and the lighting artifacts that sometimes appear on relighted images. **This is an MLBriefs article, the source code has not been reviewed!**<br> **The original source code is [[available here|https://zhhoper.github.io/dpr.html]] (last checked 2022/10/10).**

A Presentation and Short Discussion of rVAD-fast, a Fast Voice Activity Detector

Voice activity detection (VAD) usually refers to the detection of human voices in acoustic signals and is often used as a pre-processing step in numerous audio signal processing tasks. The unsupervised method proposed here was originally developed by Zheng-Hua Tan, Achintya kr. Sarkar and Najim Dehak [Computer Speech & Language, 2020] and consists of a robust segment-based approach. The voice activity detection stage follows two denoising steps. The first one detects high energy segments using a posteriori SNR weighted energy difference, and the second enhances the speech using the MSNE-mod approach. Use cases or downstream tasks include intrusion detection, speech-to-text, speaker diarization, or emotion estimation. **This is an MLBriefs article, the source code has not been reviewed!**<br> **The original source code is [[available here|https://github.com/zhenghuatan/rVAD/tree/d41f5354317bf13c1d8b31cb4f7ad4bf5112cd34]] (last checked 2022/10/10).**

A Brief Analysis of the Dense Extreme Inception Network for Edge Detection

This work describes DexiNed, a Dense Extreme Inception Network for Edge Detection proposed by Xavier Soria, Edgar Riba and Angel Sappa in [IEEE Winter Conference on Applications of Computer Vision (WACV), 2020]. The network is organized in blocks that extract edges at different resolutions, which are then merged to produce a multiscale edge map. For training, the authors introduced an annotated dataset (BIPED) specifically designed for edge detection. We perform a brief analysis of the results produced by DexiNed, highlighting its quality but also indicating its limitations. Overall, DexiNed produces state-of-the-art results. **This is an MLBriefs article, the source code has not been reviewed!**<br> **The original source code is [[available here|https://github.com/xavysp/DexiNed/tree/master/legacy/]] (last checked 2022/10/07).**

Phase Unwrapping using a Joint CNN and SQD-LSTM Network

Phase unwrapping techniques are used in various applications, including Synthetic Aperture Radar (SAR) interferometry (InSAR). Deep learning methods have been recently proposed to tackle this problem. This work aims at explaining and evaluating the method proposed by Perera et al. in [A joint convolutional and spatial quad-directional LSTM network for phase unwrapping, ICASSP 2021]. Furthermore, we provide an online demo to simulate phase images and run them through the network. The network performance can be tested visually and through metrics such as the error standard deviation. The simulation can provide some out-of-distribution data, especially with the added atmospheric signal specific to the InSAR phase. **This is an MLBriefs article, the source code has not been reviewed!**<br> **The original source code is [[available here|https://github.com/Laknath1996/DeepPhaseUnwrap/tree/9316c1b1272c0457000fbbaaff850303d923df88]] (last checked 2022/10/04).**

A Brief Analysis of the Holistically-Nested Edge Detector

This work describes the HED method for edge detection. HED uses a neural network based on a VGG16 backbone, supplemented with some extra layers for merging the results at different scales. The training was performed on an augmented version of the BSDS500 dataset. We perform a brief analysis of the results produced by HED, highlighting its quality but also indicating its limitations. Overall, HED produces state-of-the-art results. **This is an MLBriefs article, the source code has not been reviewed!**<br> **The original source code is [[available here|https://github.com/sniklaus/pytorch-hed/tree/8db09037f2491abbed4e5d7e37ec75210a2dbfbf]] (last checked 2022/10/03).**

Original source code as used in Werner et al. "Land-neutral negative emissions through biochar-based fertilization – global potentials driven by management and pyrolysis conditions" -- submitted to Mitigation and Adaptation Strategies for Global Change

Original source code as used in Werner et al. "Land-neutral negative emissions through biochar-based fertilization – global potentials driven by management and pyrolysis conditions" -- submitted to Mitigation and Adaptation Strategies for Global Change

Issues With Variability in Electronic Health Record Data About Race and Ethnicity: Descriptive Analysis of the National COVID Cohort Collaborative Data Enclave.

BackgroundThe adverse impact of COVID-19 on marginalized and under-resourced communities of color has highlighted the need for accurate, comprehensive race and ethnicity data. However, a significant technical challenge related to integrating race and ethnicity data in large, consolidated databases is the lack of consistency in how data about race and ethnicity are collected and structured by health care organizations.ObjectiveThis study aims to evaluate and describe variations in how health care systems collect and report information about the race and ethnicity of their patients and to assess how well these data are integrated when aggregated into a large clinical database.MethodsAt the time of our analysis, the National COVID Cohort Collaborative (N3C) Data Enclave contained records from 6.5 million patients contributed by 56 health care institutions. We quantified the variability in the harmonized race and ethnicity data in the N3C Data Enclave by analyzing the conformance to health care standards for such data. We conducted a descriptive analysis by comparing the harmonized data available for research purposes in the database to the original source data contributed by health care institutions. To make the comparison, we tabulated the original source codes, enumerating how many patients had been reported with each encoded value and how many distinct ways each category was reported. The nonconforming data were also cross tabulated by 3 factors: patient ethnicity, the number of data partners using each code, and which data models utilized those particular encodings. For the nonconforming data, we used an inductive approach to sort the source encodings into categories. For example, values such as “Declined” were grouped with “Refused,” and “Multiple Race” was grouped with “Two or more races” and “Multiracial.”Results“No matching concept” was the second largest harmonized concept used by the N3C to describe the race of patients in their database. In addition, 20.7% of the race data did not conform to the standard; the largest category was data that were missing. Hispanic or Latino patients were overrepresented in the nonconforming racial data, and data from American Indian or Alaska Native patients were obscured. Although only a small proportion of the source data had not been mapped to the correct concepts (0.6%), Black or African American and Hispanic/Latino patients were overrepresented in this category.ConclusionsDifferences in how race and ethnicity data are conceptualized and encoded by health care institutions can affect the quality of the data in aggregated clinical databases. The impact of data quality issues in the N3C Data Enclave was not equal across all races and ethnicities, which has the potential to introduce bias in analyses and conclusions drawn from these data. Transparency about how data have been transformed can help users make accurate analyses and inferences and eventually better guide clinical care and public policy.

Original source code as used in Werner et al., 2022 "Potential of land-neutral negative emissions through biochar sequestration" -- submitted to Earth's Future

Original source code as used in Werner et al., 2022 "Potential of land-neutral negative emissions through biochar sequestration" -- submitted to Earth's Future

Idse-HE: Hybrid embedding graph neural network for drug side effects prediction

In drug development, unexpected side effects are the main reason for the failure of candidate drug trials. Discovering potential side effects of drugsin silicocan improve the success rate of drug screening. However, most previous works extracted and utilized an effective representation of drugs from a single perspective. These methods merely considered the topological information of drug in the biological entity network, or combined the association information (e.g. knowledge graph KG) between drug and other biomarkers, or only used the chemical structure or sequence information of drug.Consequently, to jointly learn drug features from both the macroscopic biological network and the microscopic drug molecules. We propose a hybrid embedding graph neural network model named idse-HE, which integrates graph embedding module and node embedding module. idse-HE can fuse the drug chemical structure information, the drug substructure sequence information and the drug network topology information. Our model deems the final representation of drugs and side effects as two implicit factors to reconstruct the original matrix and predicts the potential side effects of drugs. In the robustness experiment, idse-HE shows stable performance in all indicators. We reproduce the baselines under the same conditions, and the experimental results indicate that idse-HE is superior to other advanced methods. Finally, we also collect evidence to confirm several real drug side effect pairs in the predicted results, which were previously regarded as negative samples. More detailed information, scientific researchers can access the user-friendly web-server of idse-HE at http://bioinfo.jcu.edu.cn/idse-HE. In this server, users can obtain the original data and source code, and will be guided to reproduce the model results.

Mitigating Computer Limitations in Replicating Numerical Simulations of a Neural Network Model With Hodgkin-Huxley-Type Neurons.

Computational experiments have been very important to numerically simulate real phenomena in several areas. Many studies in computational biology discuss the necessity to obtain numerical replicability to accomplish new investigations. However, even following well-established rules in the literature, numerical replicability is unsuccessful when it takes the computer's limitations for representing real numbers into consideration. In this study, we used a previous published recurrent network model composed by Hodgkin-Huxley-type neurons to simulate the neural activity during development. The original source code in C/C++ was carefully refactored to mitigate the lack of replicability; moreover, it was re-implemented to other programming languages/software (XPP/XPPAUT, Python and Matlab) and executed under two operating systems (Windows and Linux). The commutation and association of the input current values during the summation of the pre-synaptic activity were also analyzed. A total of 72 simulations which must obtain the same result were executed to cover these scenarios. The results were replicated when the high floating-point precision (supplied by third-party libraries) was used. However, using the default floating-point precision type, none of the results were replicated when compared with previous results. Several new procedures were proposed during the source code refactorization; they allowed replicating only a few scenarios, regardless of the language and operating system. Thus, the generated computational “errors” were the same. Even using a simple computational model, the numerical replicability was very difficult to be achieved, requiring people with computational expertise to be performed. After all, the research community must be aware that conducting analyses with numerical simulations that use real number operations can lead to different conclusions.

Reproducible Builds: Increasing the Integrity of Software Supply Chains

Although it is possible to increase confidence in free and open source software by reviewing its source code, trusting code is not the same as trusting its executable counterparts. This article examines reproducible builds, an approach that can determine whether generated binaries correspond to the original source code.

Training Opportunities in Data Science Are Welcome in ESPGHAN: A Survey on Behalf of the Special Interest Group for Basic Science and Translational Research.

To the Editor: In recent years, big data analysis has revolutionised our approach to research. Consequent trends have seen a move towards more reproducible research including the use of open access analysis tools. We surveyed ESPGHAN members to assess current utilisation of data analysis tools and investigate potential interest in developing skills in data analysis and coding. Seventy-two individuals from 14 countries representing allied health professionals (n = 6), senior clinicians (n = 26), trainees (n = 33) and researchers (n = 13) responded. Of the top five frequently used tools: Microsoft Excel (67/70), SPSS (34/70), STATA (19/70) and R (14/70), only R is free and openly available. A striking 95% of respondents would be interested in learning another programme. When asked, 95% of individuals would be interested in face to face or online training opportunities. Similarly, 96% of individuals would be interested if ESPGHAN were to provide funding to help skills development in this area. Our ESPGHAN special interest group for basic and translational research aims to pilot a programme in collaboration with the ESPGHAN Research Committee to help support skills developments in data science. Funding would include enrolment to a data science learning platform to undertake certified courses. We also envisage providing face-to-face opportunities with bioinformaticians and statisticians in biological sciences at leading centres that can offer on-site training and support. Although our survey confirms this is a current unmet need, with this initiative we aim to provide ESPGHAN members with a career-changing opportunity that will pave the way to further involvement in high-quality research in our area (Fig. 1).FIGURE 1: Summary of key survey findings. (a) Stacked bar plot showing the frequency of usage of each data analysis tool/language. Y-axis shows the number of respondents. Bars are coloured according to the frequency of usage. (b) Key showing if the languages are considered “Open Source” defined as software for which the original source code is made freely available and may be distributed and modified and “Reproducible” which is defined as analysis that provides the original data, code and software allowing others to reach the same results and conclusions as to the authors. (c) Bar chart showing response to questions regarding interest in learning new tools and in ESPGHAN funding.

Deoptfuscator: Defeating Advanced Control-Flow Obfuscation Using Android Runtime (ART)

Code obfuscation is a technique that makes it difficult for code analyzers to understand a program by transforming its structures or operations while maintaining its original functionality. Android app developers often employ obfuscation techniques to protect business logic and core algorithm inside their app against reverse engineering attacks. On the other hand, malicious app writers also use obfuscation techniques to avoid being detected by anti-malware software. If malware analysts can mitigate the code obfuscation applied to malicious apps, they can analyze and detect the malicious apps more efficiently. This paper proposes a new tool, <i>Deoptfuscator</i>, to detect obfuscated an Android app and to restore the original source codes. <i>Deoptfuscator</i> detects an app control-flow obfuscated by <i>DexGuard</i> and tries to restore the original control-flows. <i>Deoptfuscator</i> deobfuscates in two steps: it determines whether an control-flow obfuscation technique is applied and then deobfuscates the obfuscated codes. Through experiments, we analyze how similar a deobfuscated app is to the original one and show that the obfuscated app can be effectively restored to the one similar to the original. We also show that the deobfuscated apps run normally.

FSEI-GPU: GPU accelerated simulations of the fluid–structure–electrophysiology interaction in the left heart

The reliability of cardiovascular computational models depends on the accurate solution of the hemodynamics, the realistic characterization of the hyperelastic and electric properties of the tissues along with the correct description of their interaction. The resulting fluid–structure–electrophysiology interaction (FSEI) thus requires an immense computational power, usually available in large supercomputing centers, and requires long time to obtain results even if multi–CPU processors are used (MPI acceleration). In recent years, graphics processing units (GPUs) have emerged as a convenient platform for high performance computing, as they allow for considerable reductions of the time–to–solution.This approach is particularly appealing if the tool has to support medical decisions that require solutions within reduced times and possibly obtained by local computational resources. Accordingly, our multi–physics solver [1] has been ported to GPU architectures using CUDA Fortran to tackle fast and accurate hemodynamics simulations of the human heart without resorting to large–scale supercomputers. This work describes the use of CUDA to accelerate the FSEI on heterogeneous clusters, where both the CPUs and GPUs are used in synergistically with minor modifications of the original source code. The resulting GPU accelerated code solves a single heartbeat within a few hours (from three to ten depending on the grid resolution) running on premises computing facility made of few GPU cards, which can be easily installed in a medical laboratory or in a hospital, thus opening towards a systematic computational fluid dynamics (CFD) aided diagnostic.

The NAS Parallel Benchmarks for evaluating C++ parallel programming frameworks on shared-memory architectures

The NAS Parallel Benchmarks (NPB), originally implemented mostly in Fortran, is a consolidated suite containing several benchmarks extracted from Computational Fluid Dynamics (CFD) models. The benchmark suite has important characteristics such as intensive memory communications, complex data dependencies, different memory access patterns, and hardware components/sub-systems overload. Parallel programming APIs, libraries, and frameworks that are written in C++ as well as new optimizations and parallel processing techniques can benefit if NPB is made fully available in this programming language. In this paper we present NPB-CPP, a fully C++ translated version of NPB consisting of all the NPB kernels and pseudo-applications developed using OpenMP, Intel TBB, and FastFlow parallel frameworks for multicores. The design of NPB-CPP leverages the Structured Parallel Programming methodology (essentially based on parallel design patterns). We show the structure of each benchmark application in terms of composition of few patterns (notably Map and MapReduce constructs) provided by the selected C++ frameworks. The experimental evaluation shows the accuracy of NPB-CPP with respect to the original NPB source code. Furthermore, we carefully evaluate the parallel performance on three multi-core systems (Intel, IBM Power, and AMD) with different C++ compilers (gcc, icc, and clang) by discussing the performance differences in order to give to the researchers useful insights to choose the best parallel programming framework for a given type of problem.

Modeling extra-deep electromagnetic logs using a deep neural network

Modern geosteering is heavily dependent on real-time interpretation of deep electromagnetic (EM) measurements. We have developed a methodology to construct a deep neural network (DNN) model trained to reproduce a full set of extra-deep EM logs consisting of 22 measurements per logging position. The model is trained in a 1D layered environment consisting of up to seven layers with different resistivity values. A commercial simulator provided by a tool vendor is used to generate a training data set. The data set size is limited because the simulator provided by the vendor is optimized for sequential execution. Therefore, we design a training data set that embraces the geologic rules and geosteering specifics supported by the forward model. We use this data set to produce an EM simulator based on a DNN without access to the proprietary information about the EM tool configuration or the original simulator source code. Despite using a relatively small training set size, the resulting DNN forward model is quite accurate for the considered examples: a multilayer synthetic case and a section of a published historical operation from the Goliat field. The observed average evaluation time of 0.15 ms per logging position makes it also suitable for future use as part of evaluation-hungry statistical and/or Monte Carlo inversion algorithms within geosteering workflows.

Erratum: Fibers of word maps and the multiplicities of non-abelian composition factors

We discuss two corrections concerning the paper mentioned in the title. One of them pertains to a slight error in a statement from a different paper that was used. For this statement, it is explained how to fix it and why this does not affect the correctness of our results. The other is concerned with an error in the original GAP source code used to verify some of our results. This error, which was subsequently fixed, did affect the correctness of one of our main theorems, which is now given in corrected (and stronger) form.

Simulation of Quasi-Static Crack Propagation by Adaptive Finite Element Method

The finite element method (FEM) is a widely used technique in research, including but not restricted to the growth of cracks in engineering applications. However, failure to use fine meshes poses problems in modeling the singular stress field around the crack tip in the singular element region. This work aims at using the original source code program by Visual FORTRAN language to predict the crack propagation and fatigue lifetime using the adaptive dens mesh finite element method. This developed program involves the adaptive mesh generator according to the advancing front method as well as both the pre-processing and post-processing for the crack growth simulation under linear elastic fracture mechanics theory. The stress state at a crack tip is characterized by the stress intensity factor associated with the rate of crack growth. The quarter-point singular elements are constructed around the crack tip to accurately represent the singularity of this region. Under linear elastic fracture mechanics (LEFM) with an assumption in various configurations, the Paris law model was employed to evaluate mixed-mode fatigue life for two specimens under constant amplitude loading. The framework includes a progressive analysis of the stress intensity factors (SIFs), the direction of crack growth, and the estimation of fatigue life. The results of the analysis are consistent with other experimental and numerical studies in the literature for the prediction of the fatigue crack growth trajectories as well as the calculation of stress intensity factors.

PredCom: A Predictive Approach to Collecting Approximated Communication Traces

Communication traces collected from MPI applications are an important source of information for performance optimization as they can help analysts determine communication patterns and identify inefficiencies. However, their collection, especially at scale, is time consuming, since it usually requires running the complete target application on a large number of nodes. In this work, we present PredCom, a tool-chain to generate a predictive communication proxy based on information gathered from a few small scale runs, which allows us to extract approximate communication traces with an accuracy high enough for most analysis goals. For this, we combine LLVM passes on the original source code (to capture static program structure) with parameter prediction (to capture dynamic and scaling behavior). This approach drastically reduces the time needed for collecting the communication traces, even for traces on large numbers of MPI processes. We demonstrate that PredCom generates communication traces of various applications up to 1612x faster with an accuracy loss of 0.11 on average compared to the original large-scale traces, and we show that the generated traces can be used to optimize process placement.