Black-box Model Research Articles

A Grey-Box Model of a DC/DC Boost Converter for PV Energy Systems

This paper presents a grey-box model of a DC/DC boost converter for PV energy systems. The proposed model contains a white-box model part and a black-box model part together to prepare a better model for the PV boost converter. The white-box model part is used for knowledge of the circuit by mathematical equations since the black-box model part is used for unknown parameters such as temperature and electromagnetic interference. The black-box part of the proposed model is created by a nonlinear system identification of a real boost converter circuit with an artificial neural network. The precision of the mathematical model and the advantages of the fast prediction ability of the artificial neural network were used together. The proposed grey-box model is compared with the existing state-space and black-box models and experimental results. The results of the study showed that the average correlation between the proposed grey-box model output and the experimental results is 97.52%. Therefore, the proposed model can be used for analyzing DC/DC boost converter output characteristics before field applications.

Enhancing explainable SARS-CoV-2 vaccine development leveraging bee colony optimised Bi-LSTM, Bi-GRU models and bioinformatic analysis

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a single-stranded RNA virus that caused the outbreak of the coronavirus disease 2019 (COVID-19). The COVID-19 outbreak has led to millions of deaths and economic losses globally. Vaccination is the most practical solution, but finding epitopes (antigenic peptide regions) in the SARS-CoV-2 proteome is challenging, costly, and time-consuming. Here, we proposed a deep learning method based on standalone Recurrent Neural networks to predict epitopes from SARS-CoV-2 proteins easily. We optimised the standalone Bidirectional Long Short-Term Memory (Bi-LSTM) and Bidirectional Gated Recurrent Unit (Bi-GRU) with a bioinspired optimisation algorithm, namely, Bee Colony Optimization (BCO). The study shows that LSTM-based models, particularly BCO-Bi-LSTM, outperform all other models and achieve an accuracy of 0.92 and AUC of 0.944. To overcome the challenge of understanding the model predictions, explainable AI using the Shapely Additive Explanations (SHAP) method was employed to explain how Blackbox models make decisions. Finally, the predicted epitopes led to the development of a multi-epitope vaccine. The multi-epitope vaccine effectiveness evaluation is based on vaccine toxicity, allergic response risk, and antigenic and biochemical characteristics using bioinformatic tools. The developed multi-epitope vaccine is non-toxic and highly antigenic. Codon adaptation, cloning, gel electrophoresis assess genomic sequence, protein composition, expression and purification while docking and IMMSIM servers simulate interactions and immunological response, respectively. These investigations provide a conceptual framework for developing a SARS-CoV-2 vaccine.

Iterative learning robust optimization - with application to medium optimization of CHO cell cultivation in continuous monoclonal antibody production

In the presence of uncertainty, the optimum obtained based on a nominal identified model can neither provide any performance guarantee nor ensure that critical constraints are satisfied, which is crucial for e.g., bioprocess applications characterized by a high degree of complexity combined with costly experiments. Hence, uncertainty should be considered in the optimization and, furthermore, experiments designed to reduce the uncertainty most important for optimization. Herein, we propose a general framework that combines model-based robust optimization with optimal experiment design. The proposed framework can take advantage of prior knowledge in the form of a mechanistic model structure, and the importance of this is demonstrated by comparing it to more standard black-box models typically employed in learning. Through optimal experiment design, we repeatedly reduce the uncertainty most relevant for optimization so as to maximize the potential for improving the worst-case performance by balancing between exploration and exploitation. This makes the proposed method an efficient model-based robust optimization framework, especially in cases with limited experimental resources. The main part of the paper focuses on the case with modeling uncertainty that can be reduced with the availability of more experimental data. Towards the end of the paper, we consider extending the method to also include inherent uncertainty, such as input uncertainty and unmeasured disturbances. The effectiveness of the method is illustrated through a realistic simulation case study of medium optimization of Chinese hamster ovary cell cultivation in continuous monoclonal antibody production, where the metabolic network consists of 23 extracellular metabolites and 126 reactions.

On relevant features for the recurrence prediction of urothelial carcinoma of the bladder

BackgroundUrothelial bladder cancer (UBC) is characterized by a high recurrence rate, which is predicted by scoring systems. However, recent studies show the superiority of Machine Learning (ML) models. Nevertheless, these ML approaches are rarely used in medical practice because most of them are black-box models, that cannot adequately explain how a prediction is made. ObjectiveWe investigate the global feature importance of different ML models. By providing information on the most relevant features, we can facilitate the use of ML in everyday medical practice. Design, setting, and participantsThe data is provided by the cancer registry Rhineland-Palatinate gGmbH, Germany. It consists of numerical and categorical features of 1,944 patients with UBC. We retrospectively predict 2-year recurrence through ML models using Support Vector Machine, Gradient Boosting, and Artificial Neural Network. We then determine the global feature importance using performance-based Permutation Feature Importance (PFI) and variance-based Feature Importance Ranking Measure (FIRM). ResultsWe show reliable recurrence prediction of UBC with 82.02% to 83.89% F1-Score, 83.95% to 84.49% Precision, and an overall performance of 69.20% to 70.82% AUC on testing data, depending on the model. Gradient Boosting performs best among all black-box models with an average F1-Score (83.89%), AUC (70.82%), and Precision (83.95%). Furthermore, we show consistency across PFI and FIRM by identifying the same features as relevant across the different models. These features are exclusively therapeutic measures and are consistent with findings from both medical research and clinical trials. ConclusionsWe confirm the superiority of ML black-box models in predicting UBC recurrence compared to more traditional logistic regression. In addition, we present an approach that increases the explanatory power of black-box models by identifying the underlying influence of input features, thus facilitating the use of ML in clinical practice and therefore providing improved recurrence prediction through the application of black-box models.

Uncovering the heterogeneity of a solar flare mechanism with mixture models

The physics of solar flares occurring on the Sun is highly complex and far from fully understood. However, observations show that solar eruptions are associated with the intense kilogauss fields of active regions, where free energies are stored with field-aligned electric currents. With the advent of high-quality data sources such as the Geostationary Operational Environmental Satellites (GOES) and Solar Dynamics Observatory (SDO)/Helioseismic and Magnetic Imager (HMI), recent works on solar flare forecasting have been focusing on data-driven methods. In particular, black box machine learning and deep learning models are increasingly being adopted in which underlying data structures are not modeled explicitly. If the active regions indeed follow the same laws of physics, similar patterns should be shared among them, reflected by the observations. Yet, these black box models currently used in the literature do not explicitly characterize the heterogeneous nature of the solar flare data within and between active regions. In this paper, we propose two finite mixture models designed to capture the heterogeneous patterns of active regions and their associated solar flare events. With extensive numerical studies, we demonstrate the usefulness of our proposed method for both resolving the sample imbalance issue and modeling the heterogeneity for rare energetic solar flare events.

Local Interpretations for Explainable Natural Language Processing: A Survey

As the use of deep learning techniques has grown across various fields over the past decade, complaints about the opaqueness of the black-box models have increased, resulting in an increased focus on transparency in deep learning models. This work investigates various methods to improve the interpretability of deep neural networks for Natural Language Processing (NLP) tasks, including machine translation and sentiment analysis. We provide a comprehensive discussion on the definition of the term interpretability and its various aspects at the beginning of this work. The methods collected and summarised in this survey are only associated with local interpretation and are specifically divided into three categories: 1) interpreting the model’s predictions through related input features; 2) interpreting through natural language explanation; 3) probing the hidden states of models and word representations.

Relative Keys: Putting Feature Explanation into Context

Formal feature explanations strictly maintain perfect conformity but are intractable to compute, while heuristic methods are much faster but can lead to problematic explanations due to lack of conformity guarantees. We propose relative keys that have the best of both worlds. Relative keys associate feature explanations with a set of instances as context, and warrant perfect conformity over the context as formal explanations do, whilst being orders of magnitudes faster and working for complex blackbox models. Based on it, we develop CCE, a prototype that computes explanations with provably bounded conformity and succinctness, without accessing the models. We show that computing the most succinct relative keys is NP-complete and develop various algorithms for it under the batch and online models. Using 9 real-life datasets and 7 state-of-the-art explanation methods, we demonstrate that CCE explains cases where existing methods cannot, and provides more succinct explanations with perfect conformity for cases they can; moreover, it is 2 orders of magnitude faster.

Nonlinear dimensionality reduction then and now: AIMs for dissipative PDEs in the ML era

This study presents a collection of purely data-driven workflows for constructing reduced-order models (ROMs) for distributed dynamical systems. The ROMs we focus on, are data-assisted models inspired by, and templated upon, the theory of Approximate Inertial Manifolds (AIMs); the particular motivation is the so-called post-processing Galerkin method of Garcia-Archilla, Novo and Titi. Its applicability can be extended: the need for accurate truncated Galerkin projections and for deriving closed-formed corrections can be circumvented using machine learning tools. When the right latent variables are not a priori known, we illustrate how autoencoders as well as Diffusion Maps (a manifold learning scheme) can be used to discover good sets of latent variables and test their explainability. The proposed methodology can express the ROMs in terms of (a) theoretical (Fourier coefficients), (b) linear data-driven (POD modes) and/or (c) nonlinear data-driven (Diffusion Maps) coordinates. Both Black-Box and (theoretically-informed and data-corrected) Gray-Box models are described; the necessity for the latter arises when truncated Galerkin projections are so inaccurate as to not be amenable to post-processing. We use the Chafee-Infante reaction-diffusion and the Kuramoto-Sivashinsky dissipative partial differential equations to illustrate and successfully test the overall framework.

The Application of Artificial Intelligence in the Research and Development of Traditional Chinese Medicine

Review The Application of Artificial Intelligence in the Research and Development of Traditional Chinese Medicine Zhipeng Ke 1,2, Minxuan Liu 1,2,3, Jing Liu 1,2, Zhenzhen Su 1,2, Lu Li 1,2, Mengyu Qian 1,2, Xinzhuang Zhang 1,2, Tuanjie Wang 1,2, Liang Cao 1,2, Zhenzhong Wang 1,2, and Wei Xiao 1,2, * 1 National Key Laboratory on Technologies for Chinese Medicine Pharmaceutical Process Control and Intelligent Manufacture, Lianyungang 222106, China 2 Jiangsu Kanion Pharmaceutical Co., Ltd, Lianyungang 222104, China 3 ‍School of Pharmacy, Nanjing University of Chinese Medicine, Nanjing 210009, China * Correspondence: xw_kanion@163.com     Received: 4 September 2023 Accepted: 4 November 2023 Published: 6 March 2024   Abstract: With the accumulation of data in the pharmaceutical industry and the development of artificial intelligence technology, various artificial intelligence methods have been successfully employed in the drug discovery process. The integration of artificial intelligence in Traditional Chinese medicine has also gained momentum, encompassing quality control of Chinese patent medicines, prescriptions optimization, discovery of effective substances, and prediction of side effects. However, artificial intelligence also faces challenges and limitations in Traditional Chinese medicine development, such as data scarcity and complexity, lack of interdisciplinary professionals, black-box models, etc. Therefore, more research and collaboration are needed to address these issues and explore the best ways to integrate artificial intelligence and Traditional Chinese medicine to improve human health.

Explainable Classification of Patients with Primary Hyperparathyroidism Using Highly Imbalanced Clinical Data Derived from Imaging and Biochemical Procedures

Primary hyperthyroidism (PHPT) is a common endocrine disorder characterized by hypercalcemia and elevated parathyroid hormone (PTH) levels. The most common cause is a single parathyroid adenoma, though the rest of the cases are due to multiglandular disease [double adenoma/hyperplasia]. The main focus driving this work is to develop a computer-aided classification model relying on clinical data to classify PHPT instances and, at the same time, offer explainability for the classification process. A highly imbalanced dataset was created using biometric and clinical data from 134 patients (six total features, 20.2% multiglandular instances). The features used by the current study are age, sex, max diameter index, number of deficiencies, Wisconsin index, and the reference variable indicating the type of PHPT. State-of-the-art machine learning (ML) classification algorithms were used in order to create trained prediction models and give predicted classifications based on all features/indexes. Of the ML models considered (Support Vector Machines, CatBoost, LightGBM, and AdaBoost), LightGBM was able to procure the best performing prediction model. Given the highly imbalanced nature of the particular dataset, oversampling was opted for, so as to increase prediction robustness for both classes. The ML model’s performance was then evaluated using common metrics and stratified ten-fold validation. The significance of this work is rooted in two axes: firstly, in the incorporation of oversampling to smooth out the highly imbalanced dataset and offer good prediction accuracy for both classes, and secondly, in offering an explainability aspect to an otherwise black-box ML prediction model. The maximum achievable accuracy for adenoma is 86.9% and for multigland disease 81.5%. Summarizing the above, this study demonstrates the potential for an ML approach to improve the diagnosis of PHPT and also highlights the importance of explainable artificial intelligence (AI).

Improving adversarial transferability through frequency enhanced momentum

The emergence of adversarial examples seriously affects the practical security deployment of convolutional neural networks. The existing attack algorithms perform brilliantly under white-box scenarios, but they show weak transferability when faced with unknown black-box models. Recent studies have revealed that models have different interests in different frequency components of images, and the low-frequency characteristics play a non-negligible role in the decision-making of models. In this article, we present the frequency enhanced momentum iterative attack, called FE-MI-FGSM. Specifically, we use multiple convolution kernels for Gaussian filtering of the image before each gradient update to push the processed image closer to the common decision boundaries of multiple models. Then, the average gradient of the white-box model to these processed images is obtained and used as the perturbation direction to generate adversarial examples with a high success rate of white-box attack and high transferability. The empirical results show that compared with the current mainstream gradient-based methods, our method performs better on both normally trained and adversarially trained models. Besides, our method can combine with gradient-based methods which integrate convergence algorithms or input transformations for the sake of reaching satisfactory improvement of the transferability.

Comparative Study-Based Data-Driven Models for Lithium-Ion Battery State-of-Charge Estimation

Batteries have been considered a key element in several applications, ranging from grid-scale storage systems through electric vehicles to daily-use small-scale electronic devices. However, excessive charging and discharging will impair their capabilities and could cause their applications to fail catastrophically. Among several diagnostic indices, state-of-charge estimation is essential for evaluating a battery’s capabilities. Various approaches have been introduced to reach this target, including white, gray, and black box or data-driven battery models. The main objective of this work is to provide an extensive comparison of currently highly utilized machine learning-based estimation techniques. The paper thoroughly investigates these models’ architectures, computational burdens, advantages, drawbacks, and robustness validation. The evaluation’s main criteria were based on measurements recorded under various operating conditions at the Energy Systems Research Laboratory (ESRL) at FIU for the eFlex 52.8 V/5.4 kWh lithium iron phosphate battery pack. The primary outcome of this research is that, while the random forest regression (RFR) model emerges as the most effective tool for SoC estimation in lithium-ion batteries, there is potential to enhance the performance of simpler models through strategic adjustments and optimizations. Additionally, the choice of model ultimately depends on the specific requirements of the task at hand, balancing the need for accuracy with the complexity and computational resources available and how it can be merged with other SoC estimation approaches to achieve high precision.

Evaluating text classification with explainable artificial intelligence

<span lang="EN-US">Nowadays, artificial intelligence (AI) in general and machine learning techniques in particular has been widely employed in automated systems. Increasing complexity of these machine learning based systems have consequently given rise to blackbox models that are typically not understandable or explainable by humans. There is a need to understand the logic and reason behind these automated decision-making black box models as they are involved in our day-to-day activities such as driving, facial recognition identity systems, online recruitment. Explainable artificial intelligence (XAI) is an evolving field that makes it possible for humans to evaluate machine learning models for their correctness, fairness, and reliability. We extend our previous research work and perform a detailed analysis of the model created for text classification and sentiment analysis using a popular Explainable AI tool named local interpretable model agnostic explanations (LIME). The results verify that it is essential to evaluate machine learning models using explainable AI tools as accuracy and other related metrics does not ensure the correctness, fairness, and reliability of the model. We also present the comparison of explainability and interpretability of various machine learning algorithms using LIME.</span>

An innovative hybrid modeling approach for simultaneous prediction of cell culture process dynamics and product quality.

The use of hybrid models is extensively described in the literature to predict the process evolution in cell cultures. These models combine mechanistic and machine learning methods, allowing the prediction of complex process behavior, in the presence of many process variables, without the need to collect a large amount of data. Hybrid models cannot be directly used to predict final product critical quality attributes, or CQAs, because they are usually measured only at the end of the process, and more mechanistic knowledge is needed for many classes of CQAs. The historical models can instead predict the CQAs better; however, they cannot directly relate manipulated process parameters to final CQAs, as they require knowledge of the process evolution. In this work, we propose an innovative modeling approach based on combining a hybrid propagation model with a historical data-driven model, that is, the combined hybrid model, for simultaneous prediction of full process dynamics and CQAs. The performance of the combined hybrid model was evaluated on an industrial dataset and compared to classical black-box models, which directly relate manipulated process parameters to CQAs. The proposed combined hybrid model outperforms the black-box model by 33% on average in predicting the CQAs while requiring only around half of the data for model training to match performance. Thus, in terms of model accuracy and experimental costs, the combined hybrid model in this study provides a promising platform for process optimization applications.

86P Artificial intelligence tools to predict early progression of non-small cell lung cancer patients treated with immunotherapy: Preliminary evidence from a single-centre experience of black-box and white-box models pairing

86P Artificial intelligence tools to predict early progression of non-small cell lung cancer patients treated with immunotherapy: Preliminary evidence from a single-centre experience of black-box and white-box models pairing

Experimental and numerical investigation of a novel high-temperature heat pump for sensible and latent heat delivery

This paper investigates, both experimentally and numerically, the performance of a high-temperature heat pump (HTHP) prototype for steam production and thermal energy storage applications using R-1233zd(E) as a refrigerant. The main novelties of this prototype are the incorporation of an external subcooler to separate the sensible and latent heat production. Other novel points are the testing of a new variable-speed piston compressor that endures a discharge temperature of up to 160 °C, and the development of an advanced numerical model based on the components’ specifications. In this temperature range there are very few detailed studies up-to-date. The proposed HTHP was tested experimentally in 50 different working points selected based on the compressor’s limits and on the general requirements for industrial applications requiring heat at high temperatures, up to 150 °C. The results show that the proposed HTHP can deliver a total heat of 38.6 kW at 148.5 °C, within a temperature lift of 66.8 K, with an electric power consumption of 10.7 kW, and with a heating coefficient of performance (COP) of 3.6. Moreover, using the external subcooler resulted in a substantial increase in the COP, which can reach 33 % compared with similar HTHPs without a subcooler. After successfully validating the HTHP model with the experimental results, extended performance maps were developed based on more than 500 possible operating conditions covering the most common steam generation and energy storage applications. Finally, four performance correlations were obtained, with an accuracy of ±10 %, hereby providing a black-box model to ease the integration of such HTHPs into dynamic simulations of industrial processes.

Attacking Click-through Rate Predictors via Generating Realistic Fake Samples

How to construct imperceptible (realistic) fake samples is critical in adversarial attacks. Due to the sample feature diversity of a recommender system (containing both discrete and continuous features), traditional gradient-based adversarial attack methods may fail to construct realistic fake samples. Meanwhile, most recommendation models adopt click-through rate (CTR) predictors, which usually utilize black-box deep models with discrete features as input. Thus, how to efficiently construct realistic fake samples for black-box recommender systems is still full of challenges. In this article, we propose a hierarchical adversarial attack method against black-box CTR models via generating realistic fake samples, named CTRAttack. To better train the generation network, the weights of its embedding layer are shared with those of the substitute model, with both the similarity loss and classification loss used to update the generation network. To ensure that the discrete features of the generated fake samples are all real, we first adopt the similarity loss to ensure that the distribution of the generated perturbed samples is sufficiently close to the distribution of the real features, and then the nearest neighbor algorithm is used to retrieve the most appropriate features for non-existent discrete features from the candidate instance set. Extensive experiments demonstrate that CTRAttack can not only effectively attack the black-box recommender systems but also improve the robustness of these models while maintaining prediction accuracy.

Low-order gray-box modeling of heating buildings and the progressive dimension reduction identification of uncertain model parameters

Heating buildings have a huge heat capacity and have abilities to accept fluctuating heat supply from renewables. Creating a low-order scalable model that describes the thermal dynamics of coupled heat transfer, charge, and discharge processes among heating buildings is essential to unlocking the thermal storage potential of building mass and achieving space heating decarbonization. White-box models tend to be too complex and black-box models lack scalability. Thus, a low-order gray-box model with uncertain lumped parameters is proposed based on comprehensive heat transfer analyses and necessary assumptions. Thermal mass in heating buildings is divided into four independent heat storage units. The building envelope that can be heated directly or not is regarded as a different heat storage unit. To reduce the search space dimension of model identification, the decision variables are downsized to essential model parameters based on the sensitivity analysis. Moreover, a progressive dimension reduction method is introduced to further improve the probability that the optimization algorithm converges to a global minimum. Case studies show that the calibrated model with global optimal solutions of uncertain model parameters has a good performance on the prediction of the average temperature of indoor air, the average absolute error is only 0.1 °C.

Marginal effects for non-linear prediction functions

AbstractBeta coefficients for linear regression models represent the ideal form of an interpretable feature effect. However, for non-linear models such as generalized linear models, the estimated coefficients cannot be interpreted as a direct feature effect on the predicted outcome. Hence, marginal effects are typically used as approximations for feature effects, either as derivatives of the prediction function or forward differences in prediction due to changes in feature values. While marginal effects are commonly used in many scientific fields, they have not yet been adopted as a general model-agnostic interpretation method for machine learning models. This may stem from the ambiguity surrounding marginal effects and their inability to deal with the non-linearities found in black box models. We introduce a unified definition of forward marginal effects (FMEs) that includes univariate and multivariate, as well as continuous, categorical, and mixed-type features. To account for the non-linearity of prediction functions, we introduce a non-linearity measure for FMEs. Furthermore, we argue against summarizing feature effects of a non-linear prediction function in a single metric such as the average marginal effect. Instead, we propose to average homogeneous FMEs within population subgroups, which serve as conditional feature effect estimates.

Short-term trajectory prediction for individual metro passengers based on multi-level periodicity mining from semantic trajectory

Short-term trajectory prediction for individual metro passengers is an essential component of many location-based applications, such as bottleneck identification, anomalous trajectory detection, and emergency evacuation. Accurate and interpretable trajectory prediction in short-term remains challenging since the trajectory of individual metro passengers is influenced by time-varying and periodic mobility patterns. Moreover, it is still difficult for deep learning-based prediction approaches to intuitively explain the individual periodic mobility patterns captured by the black-box model. In this study, a long short-term memory method based on multi-level periodicity mining (LSTM-MP) from semantic trajectory is developed to achieve accurate and interpretable short-term trajectory prediction for individual metro passengers. The newly defined semantic trajectories of metro passengers that combine the semantic information about the geometric and operational characteristics of the metro with the raw trajectory data are constructed to help obtain fine-grained periodic mobility patterns of individual passengers. Multi-level periodicity which includes long-range, weekly-interval, and daily-interval periodicities, is calculated with variable time granularity and its polymerization from the constructed semantic trajectories. Finally, the characteristics of multi-level periodicity are considered by developing a principal trajectory attention mechanism to select the most relevant trajectories for prediction in the LSTM-MP framework. The prediction performance of LSTM-MP is compared with 2-order Markov chain and Naïve-LSTM algorithms using the large real-world anonymous passenger WiFi probe dataset. Experimental results show that LSTM-MP can predict when and where an individual will be in the future time series with an average accuracy of 82.3%. This demonstrates the effectiveness of LSTM-MP in the short-term trajectory prediction for individual metro passengers. Potential applications of the short-term trajectory prediction for individual metro passengers are also identified.