Black-box Approach Research Articles

The Multistate Quantum Monte Carlo Algebraic Diagrammatic Construction Method.

A multistate formulation of the recently developed quantum Monte Carlo (QMC) algebraic diagrammatic construction (ADC) method, QMCADC, is presented. QMCADC solves the Hermitian eigenvalue problem of the second-order ADC scheme for the polarization propagator stochastically by combining ADC schemes with projector quantum Monte Carlo (PQMC). It allows for massively parallel distributed computing and exploits the sparsity of the effective ADC matrix, thereby relaxing memory and processing requirements of ADC methods significantly. Here, the theory and implementation of the multistate variant of QMCADC are described, and our first proof-of-principle calculations for various molecular systems are shown. Indeed, multistate QMCADC enables sampling of an arbitrary number of low-lying excited states and can reproduce their vertical excitation energies with a marginal controllable error. The performance of multistate QMCADC is examined in terms of state-wise and overall accuracy as well as with respect to the balance in the treatments of excited states relatively to each other. The results are very promising as they show bias and imbalances among excited states to diminish as the number of sampling points increases. Furthermore, the impact of the quality of trial wave functions on the vertical excitation energies is investigated. A black-box approach for the generation of high quality trial wave functions internally is given.

Comparison of energy-efficiency benchmarking methodologies for residential buildings

Benchmarking with defined metrics is an excellent tool to track progress and obtain the desired goals. Benchmarking end-use energy in the residential sector has been a topic of great interest among researchers in recent times. This research analyses the buildings’ energy performance based on a devised benchmarking procedure for residential buildings in the Indian city of Jaipur. Creation of benchmarking system involves collection, analysis and verification of the results obtained from a similar set of data into consideration. This study explores the relationship between the Energy Performance Index (EPI) and with other influencing factors like; Area, end-use appliances (Fridge, AC, Cooler, etc.) to ensure the contribution of each variable towards the buildings’ end-use energy consumption. In this study, information has been collected and analysed from over 2700 houses of Jaipur City. The independent and dependent variables have been identified, and a five-star benchmarking framework has been developed. There are very few studies available investigating the effectiveness of various benchmarking techniques, the domain still lacks a comparative performance study of black-box and gray-box benchmarking techniques. This study has selected two black box approaches and one gray box approach to design a benchmarking model and has compared the energy performance of sample buildings. According to the Pearson and Spearman correlation coefficients, Multiple Linear Regression (MLR) and Bayesian ranking scores are the most consistent for energy benchmarking of the residential building sector. This study proposes a novel implementation of the composite indicator (C.I.) as a platform for designing energy benchmarking tables. Finally, the study concludes with recommendations for future work to employ numerous or hybrid combinations of benchmarking methodologies that are expected to provide a more accurate depiction of the energy performance of buildings.

An explainable and interpretable model for attention deficit hyperactivity disorder in children using EEG signals

Background:Attention deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder that affects a person’s sleep, mood, anxiety, and learning. Early diagnosis and timely medication can help individuals with ADHD perform daily tasks without difficulty. Electroencephalogram (EEG) signals can help neurologists to detect ADHD by examining the changes occurring in it. The EEG signals are complex, non-linear, and non-stationary. It is difficult to find the subtle differences between ADHD and healthy control EEG signals visually. Also, making decisions from existing machine learning (ML) models do not guarantee similar performance (unreliable). Method:The paper explores a combination of variational mode decomposition (VMD), and Hilbert transform (HT) called VMD-HT to extract hidden information from EEG signals. Forty-one statistical parameters extracted from the absolute value of analytical mode functions (AMF) have been classified using the explainable boosted machine (EBM) model. The interpretability of the model is tested using statistical analysis and performance measurement. The importance of the features, channels and brain regions has been identified using the glass-box and black-box approach. The model’s local and global explainability has been visualized using Local Interpretable Model-agnostic Explanations (LIME), SHapley Additive exPlanations (SHAP), Partial Dependence Plot (PDP), and Morris sensitivity. To the best of our knowledge, this is the first work that explores the explainability of the model prediction in ADHD detection, particularly for children. Results:Our results show that the explainable model has provided an accuracy of 99.81%, a sensitivity of 99.78%, 99.84% specificity, an F-1 measure of 99.83%, the precision of 99.87%, a false detection rate of 0.13%, and Mathew’s correlation coefficient, negative predicted value, and critical success index of 99.61%, 99.73%, and 99.66%, respectively in detecting the ADHD automatically with ten-fold cross-validation. The model has provided an area under the curve of 100% while the detection rate of 99.87% and 99.73% has been obtained for ADHD and HC, respectively. Conclusions:The model show that the interpretability and explainability of frontal region is highest compared to pre-frontal, central, parietal, occipital, and temporal regions. Our findings has provided important insight into the developed model which is highly reliable, robust, interpretable, and explainable for the clinicians to detect ADHD in children. Early and rapid ADHD diagnosis using robust explainable technologies may reduce the cost of treatment and lessen the number of patients undergoing lengthy diagnosis procedures.

Structure and performance of fully connected neural networks: Emerging complex network properties

Understanding the behavior of Artificial Neural Networks is one of the main topics in the field recently, as black-box approaches have become usual since the widespread of deep learning. Such high-dimensional models may manifest instabilities and weird properties that resemble complex systems. Therefore, we propose Complex Network (CN) techniques to analyze the structure and performance of fully connected neural networks. For that, we build a dataset with 4 thousand models (varying the initialization seed) and their respective CN properties. This is the first work to explore the CN properties of an ample number of fully connected networks accounting for the variance caused by random weight initialization. The networks are trained in a supervised classification setup considering four vision benchmarks and then approached as a weighted and undirected graph of neurons and synapses (learned weights). Results show that neuronal centrality is highly correlated to network classification performance. We also propose the concept of Bag-Of-Neurons (BoN), a CN-based approach for finding topological signatures linking similar neurons. Results suggest that six neuronal types emerge in such networks, independently of the target domain, and are distributed differently according to classification accuracy. We also tackle specific CN properties related to performance, such as higher subgraph centrality on lower-performing models. Our findings suggest that CN properties play a critical role in the performance of fully connected neural networks, with topological patterns emerging independently on a wide range of models.

Reliable adaptive distributed hyperparameter optimization (RadHPO) for deep learning training and uncertainty estimation

Training and validation of Neural Networks (NN) are very computationally intensive. In this paper, we propose a distributed system based NN infrastructure that achieves two goals: to accelerate model training, specifically for hyperparameter optimization, and to re-use some of these intermediate models to evaluate the uncertainty of the model. By accelerating model training, we can obtain a large set of potential models and compare them in a shorter amount of time. Automating this process reduces development time and provides an easy way to compare different models. Our application runs different models on distinct servers with a single training data set, each with tweaked hyperparameters. By adding uncertainty to our results, our framework provides not just a single prediction but a distribution over predictions. Adding uncertainty is essential to some NN applications since most models assume that the input data distributions are identical between test and validation. However, in reality, they are producing some catastrophic mistakes. Since our solution is a distributed system, we make our implementation robust to common distributed system failures (servers going down, loss of communication among some nodes, and others). Furthermore, we use a gossip-style heartbeat protocol for failure detection and recovery. Finally, some preliminary results using a black-box approach to generate the training models show that our infrastructure scales well in different hardware platforms.

AdVulCode: Generating Adversarial Vulnerable Code against Deep Learning-Based Vulnerability Detectors

Deep learning-based vulnerability detection models have received widespread attention; however, these models are susceptible to adversarial attack, and adversarial examples are a primary research direction to improve the robustness of the models. There are three main categories of adversarial example generation methods for source code tasks: changing identifier names, adding dead code, and changing code structure. However, these methods cannot be directly applied to vulnerability detection. Therefore, we propose the first study of adversarial attack on vulnerability detection models. Specifically, we utilize equivalent transformations to generate candidate statements and introduce an improved Monte Carlo tree search algorithm to guide the selection of candidate statements to generate adversarial examples. In addition, we devise a black-box approach that can be applied to widespread vulnerability detection models. The experimental results show that our approach achieves attack success rates of 16.48%, 27.92%, and 65.20%, respectively, in three vulnerability detection models with different levels of granularity. Compared with the state-of-the-art source code attack method ALERT, our method can handle models with identifier name mapping, and our attack success rate is 27.59% higher on average than ALERT.

DESIGN AND DEVELOPMENT OF ANDROID-BASED TRADITIONAL INDONESIAN CLOTHING IMAGE GUESSING GAME

In terms of cultural traditions, Indonesia is home to a wide variety of very diverse practices. As a result of different regions of the country, Indonesia is home to diverse cultural traditions. Traditional clothing is an important aspect of the Indonesian culture. The traditions and culture of each region are reflected in the clothes worn at traditional ceremonies; in addition, traditional clothing is a decisive feature of the culture of each region. Concern for the maintenance of national culture, which relates to the wearing of indigenous clothing, has been significantly reduced because of the passage of time. For example, one of the main reasons why people's interest in traditional clothing is not as strong as it used to be is because there is not enough socialization and presentation of Indonesian culture in society. The goal of this study project is to create a learning app for Android based on the guesswork of traditional Indonesian clothing. Research and Development techniques were selected as the proximity used for this investigation. Research and Development As a consequence of this research, an educational game that allows users to predict photos of traditional clothing from Indonesia will be produced on the Android platform. This game will also provide information to players about traditional musical instruments that can be found in Indonesia. It is hoped that educational video games will help in the process of exposing children to pre-existing cultures, and such games can also be seen as offering new perspectives to children. It is possible to draw the conclusion, based on the findings of tests conducted using the black box approach, that this application is capable of functioning in an appropriate way. Moreover, it can provide android smartphone customers with instructional games.

Toward Better Analysis of Breast Cancer Diagnosis: Interpretable AI for Breast Cancer Classification

Recently, some countries have been distressing with the increasing number of breast cancer cases. Those cases were extremely increased in every year. Practicaly, the increasing number of patients was caused by the manual examination. Recently, some researchers have been done in the development of AI method for solving this problem. However, AI itself still has limitation since it worked in the black-box approach which was difficult to be trusted. Thus, to overcome those problems, we proposed a method that was able to classify breast ultrasound images into two classes (benign and malignant) and able to explain how the prediction was made. Our proposed method consisted of four processes i.e., pre-processing step, development of CNN model, interpretable step and evaluation. In this research work, our proposed method performed into 780 breast ultrasound images divided into three classes (133 normal, 210 malignant, and 437 benign). In the training process, our proposed method obtained training accuracy of 0.9795, training loss of 0.0675. The validation process obtained validation accuracy of 0.8000 and validation loss of 0.5096. While, in the testing process, our proposed method achieved accuracy of 0.7923. In the interpretable process using LIME, the LIME result is covered by doctor visualization. It was indicated that LIME was suitable enough in visualizing the important features of breast cancer severity. Regarding to the results, our proposed method has a potensial to be implemented as an early detection method for classifying malignancy of breast cancer in order to help the doctor in the screening process

Explaining the black-box smoothly-A counterfactual approach.

We propose a BlackBox Counterfactual Explainer, designed to explain image classification models for medical applications. Classical approaches (e.g.,, saliency maps) that assess feature importance do not explain how imaging features in important anatomical regions are relevant to the classification decision. Such reasoning is crucial for transparent decision-making in healthcare applications. Our framework explains the decision for a target class by gradually exaggerating the semantic effect of the class in a query image. We adopted a Generative Adversarial Network (GAN) to generate a progressive set of perturbations to a query image, such that the classification decision changes from its original class to its negation. Our proposed loss function preserves essential details (e.g.,support devices) in the generated images. We used counterfactual explanations from our framework to audit a classifier trained on a chest X-ray dataset with multiple labels. Clinical evaluation of model explanations is a challenging task. We proposed clinically-relevant quantitative metrics such as cardiothoracic ratio and the score of a healthy costophrenic recess to evaluate our explanations. We used these metrics to quantify the counterfactual changes between the populations with negative and positive decisions for a diagnosis by the given classifier. We conducted a human-grounded experiment with diagnostic radiology residents to compare different styles of explanations (no explanation, saliency map, cycleGAN explanation, and our counterfactual explanation) by evaluating different aspects of explanations: (1) understandability, (2) classifier's decision justification, (3) visual quality, (d) identity preservation, and (5) overall helpfulness of an explanation to the users. Our results show that our counterfactual explanation was the only explanation method that significantly improved the users' understanding of the classifier's decision compared to the no-explanation baseline. Our metrics established a benchmark for evaluating model explanation methods in medical images. Our explanations revealed that the classifier relied on clinically relevant radiographic features for its diagnostic decisions, thus making its decision-making process more transparent to the end-user.

Parameter Estimation of a Grid-Tied Inverter Using In Situ Pseudo-Random Perturbation Sources

Inverters are playing an increasingly important role in the electrical utility grid due to the proliferation of renewable energy sources. Obtaining inverter models with accurate parameters is, therefore, essential for grid studies and design. In this paper, a methodology to estimate the output impedance and parameters of a residential grid-tied inverter is proposed. The methodology is first verified through simulation. A sensitivity analysis is conducted to determine the influence of the filter and controller parameters on the output impedance of the inverter. The simulated output impedance, voltage, and current are used in a parameter estimation methodology to obtain filter and controller parameters. It is shown that up to seven parameters can be estimated accurately. The proposed methodology is further investigated through a practical experiment. Two perturbation sources, the pseudo-random binary sequence perturbation and pseudo-random impulse sequence perturbation, are used, in turn, to perturb a residential grid-tied inverter that delivers up to 1.6 kW with the aim of obtaining its output impedance. The output impedances obtained through both pseudo-random sources are compared. It is shown that a pseudo-random binary sequence perturbation source applied in series between the grid and the inverter under test allows for the best estimation of the grid-tied inverter’s output impedance. A black-box modeling approach aimed at estimating an analytical transfer function of the output impedance from experimental data is also discussed.

A Highly Accurate Internet-Based Fake Information Detection Tool for Indonesian Twitter

The dissemination of fake information through social media has several harmful effects on the social life of a nation. Indonesia has been afflicted by the dissemination of erroneous information regarding the negative health consequences of vaccination, resulting in widespread unwillingness to undergo immunization. Therefore, it is necessary to combat such misleading information. We developed a web application using machine learning technologies to identify bogus information flowing on Indonesian Twitter. A Passive-Aggressive Classifier and n-gram tokenization are used to handle data. The application test results indicate that the detection accuracy, precision, and recall for 1-3 grams of tokenization are higher than 90%. In addition, the black box approach yields reliable findings for all application functionalities.

Special issue on “Optimal design and operation of energy systems”

Special issue on “Optimal design and operation of energy systems”

The importance of possessing knowledge on black-box components: The case of smartphone OEMs

Traditionally, the black-box approach in buyer-supplier relationships does not require knowledge exchanges between the actors, since the purchased module is autonomously developed by the supplier. However, the purpose of the article is to demonstrate that in high-tech sectors, despite the presence of black-box approaches, the OEM can achieve higher market performance if it develops in-house competences on purchased modules. To measure OEM component knowledge, a content analysis of patents in the smartphone industry is performed. The results show the importance for OEMs to accumulate both general and specific component knowledge, although modules are outsourced.

Ship Manoeuvering Modelling with a Physics-Oriented Neural Network-Based Approach

This paper proposes a novel system identification schema to obtain a model of a ship manouvering process using artificial Neural Networks (ANNs), computational models that have changed the research paradigm, bringing remarkable advantages in several fields. The ANNs capabilities in modelling nonlinear dynamical systems are undeniable and several approaches have been proposed in recent years. In our work, an ordinary black-box approach used to determine the input-output relationship of the system under investigation is initially outlined. Then, a physics-oriented approach to train a recurrent neural network structure is provided and thoroughly explained, investigating also the possible adoption of a simpler network structure. The results obtained with the physics-oriented approach in modelling the ship manouvering process are significantly better than the ones achieved with the pure black-box approach. Consequently, the physics-oriented approach resulted to be an exceptional tool for inferring the physical laws behind a nonlinear system accounting for a limited amount of data. The effectiveness of this method motivates further studies to evaluate its possible implementation in model-based control algorithms.

Model-based and Koopman-based predictive control: a braking control systems comparison

Anti-locking Braking systems are crucial safety systems in modern vehicles. In this work, we investigate the possibility to use Model Predictive Control (MPC) for braking systems by considering three different models identified from data. Specifically, we consider two models, whose structure and the identification procedure are driven by physics principles, and a third black-box modeling approach that relies on Koopman theory. By comparing the effectiveness of the three resulting MPC schemes in a high-fidelity simulation environment, we show that Koopman-based MPC can generally be a viable solution for the design of braking controllers, which might not be the case of nonlinear MPC or approximated scheme like the second one we test.

The Best Protection is Attack: Fooling Scene Text Recognition With Minimal Pixels

Scene text recognition (STR) has witnessed tremendous progress in the era of deep learning, but it also raises concerns about privacy infringement as scene texts usually contain valuable or sensitive information. Previous works in privacy protection of scene texts mainly focus on masking out the texts from the image/video. In this work, we learn from the idea of adversarial examples and use minimal pixel perturbation to protect the privacy of text information. Although there are well-established attacking methods on non-sequential vision tasks (e.g., classification), the attack on sequential tasks (e.g., scene text recognition) has not received sufficient attention yet. Moreover, existing works mainly focus on the white-box setting, which requires complete knowledge of the target model (e.g., architecture, parameters, or gradients). These requirements limit the scope of applications for the white-box adversarial attack. Therefore, we propose a novel black-box attacking approach for the STR models, only requiring prior knowledge of the model output. Besides, instead of disturbing most pixels as in existing STR attack methods, our proposed approach only manipulates a few pixels, meaning the perturbation is more inconspicuous. To determine the location and value of the manipulated pixels, we also provide an efficient Adaptive-Discrete Differential Evolution (AD <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> E) by narrowing down the continuous searching space to a discrete space. It can greatly reduce the queries to the target model. Experiments on several real-world benchmarks show the effectiveness of our proposed approach. Especially, when attacking the commercial STR engine, Baidu-OCR, our method achieves higher attack success rates by a large margin than existing approaches. Our work establishes an important step towards using the black-box adversarial attack with minimal pixels to protect the privacy of text information from being easily obtained by STR models.

Physics-Informed Hybrid GRU Neural Networks for MPC Prediction

This work describes a novel Physics-Informed Hybrid Neural Network (PIHNN) model structure based on the Gated Recurrent Unit (GRU) neural network that combines the first principle and black-box data-driven approaches. We recommend the presented modeling method when the measurement of process variables is possible, but only in some limited neighborhood of the operating points, and the available first principle model describing the process is generally correct but not accurate. We discuss three methods of data fusion. Two of them utilize the fuzzy approach, while the third one relies on an additional neural network of the Multi-Layer Perceptron (MLP) type. The first method assumes a simplified data fusion block, while the next two use machine learning to minimize the overall model error. The efficiency of the presented approach and data fusion techniques is demonstrated for a benchmark chemical process (the pH reactor). Finally, we consider applying the PIHNN model in Model Predictive Control (MPC). It gives much better control quality than the MPC controller relying on the entirely black-box GRU model.

Comparing Individualized Survival Predictions From Random Survival Forests and Multistate Models in the Presence of Missing Data: A Case Study of Patients With Oropharyngeal Cancer.

In recent years, interest in prognostic calculators for predicting patient health outcomes has grown with the popularity of personalized medicine. These calculators, which can inform treatment decisions, employ many different methods, each of which has advantages and disadvantages. We present a comparison of a multistate model (MSM) and a random survival forest (RSF) through a case study of prognostic predictions for patients with oropharyngeal squamous cell carcinoma. The MSM is highly structured and takes into account some aspects of the clinical context and knowledge about oropharyngeal cancer, while the RSF can be thought of as a black-box non-parametric approach. Key in this comparison are the high rate of missing values within these data and the different approaches used by the MSM and RSF to handle missingness. We compare the accuracy (discrimination and calibration) of survival probabilities predicted by both approaches and use simulation studies to better understand how predictive accuracy is influenced by the approach to (1) handling missing data and (2) modeling structural/disease progression information present in the data. We conclude that both approaches have similar predictive accuracy, with a slight advantage going to the MSM. Although the MSM shows slightly better predictive ability than the RSF, consideration of other differences are key when selecting the best approach for addressing a specific research question. These key differences include the methods' ability to incorporate domain knowledge, and their ability to handle missing data as well as their interpretability, and ease of implementation. Ultimately, selecting the statistical method that has the most potential to aid in clinical decisions requires thoughtful consideration of the specific goals.

Black-Box Attack using Adversarial Examples: A New Method of Improving Transferability

Adversarial examples (AEs) are malicious test-data samples (typically images) generated by applying carefully calculated perturbations to clean samples. The added perturbations are usually human-imperceptible but the AEs can fool a machine learning (ML) model to make misclassifications. Although multiple methods were proposed to generate AEs, the ability to generalize is very limited; that is, they easily overfit to their source, single, white-box ML models and the generated AEs rarely work for other models. In this paper, we propose a black-box attack approach that crafts transferable AEs that can attack a wide range of ML models without knowing those model details. Our novel method consists of an elastic momentum (EM) that expedites gradient descent to avoid early overfitting, and a random erasure (RE) technique that increases the diversity of perturbations and reduces gradient fluctuations. Our method can be applied to any gradient-based attacks to make those attacks become more transferable. We evaluate our proposed method by attacking seven state-of-the-art (SOTA) deep learning models and comparing against five SOTA attacks; we also attack nine advanced defense mechanisms that are integrated into the above models. Our results demonstrate significant improvement on the attack success rate (ASR) and transferability when using our method alone, and that it can also be easily applied to other baseline methods (which are gradient-based) to substantially improve their performance.

AdapTV: A Model-Based Test Adaptation Approach for End-to-End User Interface Testing of Smart TVs

We introduce a model-based feedback-driven test adaptation approach for end-to-end user interface testing of smart TVs. From the perspective of the TV software, the proposed approach is a non-intrusive and completely black-box approach, which operates by interpreting the screen images. Given a test suite, which is known to work in an older version of the TV, and a new version of the TV, to which the test suite should be adapted, the proposed approach first automatically discovers user interface models for both the older and the new version of TV by opportunistically crawling the TVs. Then, each test case in the test suite is executed on the old version, and the path traversed by the test case in the respective UI model is found. Finally, a semantically equivalent path in the UI model discovered for the new version of the TV is determined and dynamically executed on the new version in a feedback-driven manner. We empirically evaluate the proposed approach in a setup that closely mimics the industrial setup used by a large consumer electronics company. While the proposed approach successfully adapted all the test cases, the alternative approaches used in the experiments could not adapt any of them.