Original Set Research Articles

Effects of Injection Pressure on Performance and Emission Characteristics of CI Engine using Waste Cooking Oil (WCO) Blend

For a long period of time, vegetable oil cannot be used directly in a direct injection diesel engine. The tolerance tests may indicate significant problems. The outcome of Injection Pressures (IP) on the performance and emissions of a diesel engine powered by waste cooking oil biodiesel were explored in this study. It is investigated and standard diesel results are compared to the performance characteristics and emission studies of a single cylinder, four-stroke, direct-injection diesel engine fueled with used cooking oil in 20% (on a mass basis) blends. This study established the appropriateness of using cooking oil. The experiment consisted of running at a constant speed of 1500 rpm and then loaded gradually. The tests were conducted at 5 different loading are 0%, 25%, 50%, 75%, and 100% of the load in kW, with compression ratios of 17.5:1. The result of adjusting the injection pressure to 185–235 bars with a gap of 25 bars, with the original IP set at 210 bar. Fuel injection pressure is critical in improving engine performance and emission characteristics. The diesel engine ran on Waste Cooking Oil (WCO) at various injection pressures, including 185 bars, 210 bars, and 235 bars. The engine tests were carried out to examine how a diesel engine operating on a WCO20 biodiesel blend performed and produced emissions at varied injection pressures. Except for NOx, biodiesel blends at 235 bars injection pressure performed better and had lower emissions than those at 185 bars injection pressure. Without any modifications, the optimal fuel blend can be considered for a compression ignition engine.

Defense against adversarial attacks via textual embeddings based on semantic associative field

AbstractDeep neural networks are known to be vulnerable to various types of adversarial attacks, especially word-level attacks, in the field of natural language processing. In recent years, various defense methods are proposed against word-level attacks; however, most of those defense methods only focus on synonyms substitution-based attacks, while word-level attacks are not based on synonym substitution. In this paper, we propose a textual adversarial defense method against word-level adversarial attacks via textual embedding based on the semantic associative field. More specifically, we analyze the reasons why humans can read and understand textual adversarial examples and observe two crucial points: (1) There must be a relation between the original word and the perturbed word or token. (2) Such a kind of relation enables humans to infer original words, while humans have the ability to associations. Motivated by this, we introduce the concept of semantic associative field and propose a new defense method by building a robust word embedding, that is, we calculate the word vector by exerting the related word vector to it with potential function and weighted embedding sampling for simulating the semantic influence between words in same semantic field. We conduct comprehensive experiments and demonstrate that the models using the proposed method can achieve higher accuracy than the baseline defense methods under various adversarial attacks or original testing sets. Moreover, the proposed method is more universal, while it is irrelevant to model structure and will not affect the efficiency of training.

Visual–inertial object tracking: Incorporating camera pose into motion models

Visual object tracking for autonomy of aerial robots could become challenging especially in the presence of target or camera fast motions and long-term occlusions. This paper presents a visual-inertial tracking paradigm by incorporating the camera kinematics states into the visual object tracking pipelines. We gathered a dataset of image sequences with the addition of camera’s position and orientation measurements as well as the object’s position measurement. For the cases of long-term object occlusion, we provide ground-truth boxes derived from mapping the measured object position onto the image frame. A search zone proposal method is developed based on the estimation of object future position represented in the inertial frame and projected back onto the image frame using the camera states. This search zone, which is robust to fast camera/target motions, is fused into the original search zone settings of the base tracker. Also proposed is a measure indicating the confidence of a tracking structure in keeping track of a correct target and reporting the tracking failure in-time. Accordingly, the model updating mechanism of base tracker is modulated to avoid recovering of wrong objects as the target. The proposed modifications are benchmarked on nine visual object tracking algorithms including five state-of-art deep learning structures , namely DiMP, PrDiMP, KYS, ToMP, and MixFormer. Most of the trackers are remarkably improved by the modifications with up to 8% increase in precision. Modified PrDiMP tracker yields the best precision of 68.4%, more than all considered original (and modified) trackers. Source code and dataset are made available online.1

Multi-class granular approximation by means of disjoint and adjacent fuzzy granules

In granular computing, fuzzy sets can be approximated by granularly representable sets that are as close as possible to the original fuzzy set w.r.t. a given closeness measure. Such sets are called granular approximations. In this article, we introduce the concepts of disjoint and adjacent granules and we examine how the new definitions affect the granular approximations. First, we show that the new concepts are important for binary classification problems since they help to keep decision regions separated (disjoint granules) and at the same time to cover as much as possible of the attribute space (adjacent granules). Later, we consider granular approximations for multi-class classification problems leading to the definition of a multi-class granular approximation. Finally, we show how to efficiently calculate multi-class granular approximations for Łukasiewicz fuzzy connectives. We also provide graphical illustrations for a better understanding of the introduced concepts.

The original set designs for Bodas de sangre and Yerma: written and built spaces

Federico García Lorca (1898-1936) was one of the most influential writers of Spanish literature of the 20th century. His particular point of view, influenced by rural and urban experiences, inspired set designs with high aesthetic and compositional value. This research analyzes the original stagings of Bodas de sangre in the Teatro Beatriz (Madrid, 1933) and Yerma in the Teatro Español (Madrid, 1934). Compiling documentary material and preparing plans and technical models offer a complete view of each proposal, considering the design criteria and construction processes of those designs for which information is available today. Studying the set designers responsible for each project is also relevant. Manuel Fontanals, Santiago Ontañón, and Siegfried Burmann were leading figures in Spanish and foreign theatrical plastic art during the Second Spanish Republic and the following years. In summary, the study shows how the designs for each scenic space were created, analyzing both their compositional and technical details and the transformation process from the written space to the built one. After all, set design is a discipline belonging to ephemeral architecture, which aims to generate atmospheres and structures with a short useful life but with aesthetic, functional, and stability requirements.

Comparison evaluation of unet-based models with noise augmentation for breast cancer segmentation on ultrasound image

This paper investigates the application of convolutional neural networks (CNNs), particularly the UNet model architecture, to improve the accuracy of breast cancer tumor segmentation in ultrasound images. Accurate identification of breast cancer is essential for effective patient treatment. However, ultrasound images, often contain noise and artifacts, which can complicate the task of tumor segmentation. Therefore, to highlight the most robust architecture, modifications were made to the original set, including the addition of noise and fuzziness. In this study, a comparative study of five different variants of UNet models (UNet, Attention UNet, UNet++, Dense Inception UNet and Residual UNet) was conducted on a diverse set of ultrasound images with different breast tumors. Using consistent training methods and techniques of augmentation and adding noise to the data, an improvement in segmentation accuracy was highlighted when using the Dense Inception UNet architecture. The results have potential practical applications in the field of medical diagnosis and can assist medical professionals in automatic tumor segmentation in breast cancer ultrasound images. This study highlights the improvement of segmentation accuracy by introducing dense induction into the UNet architecture. Importantly, the Dice coefficient, a key segmentation metric, improved markedly, increasing from 0.973 to 0.976 after data augmentation. The results of the study offer promise to the medical community by offering a more accurate and reliable approach to segmenting breast cancer lesions on ultrasound images. The findings can be implemented in clinical practice to assist radiologists in early cancer diagnosis.

Modeling of bioprocess pre-stages for optimization of perfusion profiles and increased process understanding.

Improving bioprocess efficiency is important to reduce the current costs of biologics on the market, bring them faster to the market, and to improve the environmental footprint. The process intensification efforts were historically focused on the main stage, while intensification of pre-stages has started to gain attention only in the past decade. Performing bioprocess pre-stages in the perfusion mode is one of the most efficient options to achieve higher viable cell densities over traditional batch methods. While the perfusion-mode operation allows to reach higher viable cell densities, it also consumes large amount of medium, making it cost-intensive. The change of perfusion rate during a process (perfusion profile) determines how much medium is consumed, thereby running a process in optimal conditions is key to reduce medium consumption. However, the selection of the perfusion profile is often made empirically, without full understanding of bioprocess dynamics. This fact is hindering potential process improvements and means for cost reduction. In this study, we propose a process modeling approach to identify the optimal perfusion profile during bioprocess pre-stages. The developed process model was used internally during process development. We could reduce perfused medium volume by 25%-45% (project-dependent), while keeping the difference in the final cell within 5%-10% compared to the original settings. Additionally, the model helps to reduce the experimental workload by 30%-70% and to predict an optimal perfusion profile when process conditions need to be changed (e.g., higher seeding density, change of operating mode from batch to perfusion, etc.). This study demonstrates the potential of process modeling as a powerful tool for optimizing bioprocess pre-stages and thereby guiding process development, improving overall bioprocess efficiency, and reducing operational costs, while strongly reducing the need for wet-lab experiments.

Can the observable solar activity spectrum be reproduced by a simple dynamo model?

The temporal spectrum of the solar activity is more than just the main cycle. It contains different timescales, which can be considered as continuous components of the activity spectrum. The possibility of finding a realistic spectrum of the solar magnetic activity variation is analysed for several versions of a simple model for solar activity based on the original idea of E. Parker. In particular, we study the original set of partial differential equations with two versions of suppression of the dynamo action and the fourth-order dynamical system obtained by truncating the Parker equations. We show that the effects included in the models, i.e. the nonlinear dynamo suppression and the dynamical chaos, as well as random fluctuations of the dynamo drivers, are quite sufficient to obtain the main solar cycle and the continuous components of the spectrum similar to the observed ones. However, the capabilities of the approach under consideration substantially vary from one model to another. Each model reproduces a continuous component of the spectrum in a specific parameter range. This study has confirmed the view that the examination of various solar dynamo models with the aim to find a reasonable combination of main activity cycle and continuous spectrum of solar activity can be used as an additional test of their validity.

Uncertainty Quantification for the Noah‐MP Land Surface Model: A Case Study in a Grassland and Sandy Soil Region

AbstractLand surface models rely on a multitude of parameters to simulate land‐atmosphere interactions, but the parameter uncertainty can limit the reliability of model predictions. This study utilizes a parameter uncertainty quantification (UQ) framework to quantify and reduce the parameter uncertainty of the Noah‐MP land surface model in a grassland and sandy soil region in the Midwest of the USA. First, the sparse polynomial chaos expansion method which can quantify the interaction effect of parameters, is employed. A relatively small parameter sample size (i.e., 20 times of the number of parameters) was sufficient to identify the sensitive parameters; an additional sensitive parameter, the saturated soil hydraulic conductivity, was screened out compared to previous study. Then, based on the selected sensitive parameters, the weighted multi‐objective adaptive surrogate modeling‐based optimization algorithm is used as the parameter optimization method. The optimization results showed that the root mean square error of flux of latent heat (FLH) on about 82% of the total grids was reduced, and the number was about 57% for gross primary production (GPP) compared to the results using the original parameter settings, indicating that the Pareto parameter set by the UQ framework improved the Noah‐MP model in simulating FLH and GPP in a grassland and sandy soil region in the Midwest of the USA.

Assessment of Academic ESL Writing in an Online Tutorial for Graduate Students

English-as-a-second language (L2) graduate students often face challenges in developing academic writing skills, which can be exacerbated by a lack of timely support at institutional level. To help address this concern, an original set of online academic writing tutorials with a focus on the genre of Literature Review was designed by the authors to assist international graduate students enrolled in graduate programs in Canada. This paper introduces and compares multiple assessment tools employed in the online tutorial set for international English L2 graduate students. Our project pursues two major goals: First, to address the above-identified gap in Academic Writing support to English L2 graduate students with minimal costs and faculty involvement through designing the online tutorial set; second, to contribute to the research on e-learning of Academic English as a second language (ESL) writing in terms of developing integrated tools for online tutorial building, analysis of texts produced by learners, and assessment of learners’ writing progress. The research questions are: 1. What resources can be combined to develop an online tutorial set for graduate students at minimal costs? 2. How can the learners’ progress in academic ESL writing be assessed with different assessment tools? First, the paper describes the tools employed in the tutorial construction: MoodleCloud platform, H5P interactive e-book designed by the authors, surveys, and assessment tools. Second, we present and compare assessment tools employed to evaluate learners’ writing progress: Expert assessment with an analytic rubric, self-assessments of progress by the participants, and automated text analysis with corpus-based tools as reported in Li, Makarova, and Wang (2023). A comparison of the scores across the three assessment tools shows some discrepancies, which seems to suggest that combined tools yield a more comprehensive picture. The expert assessments and self-assessments demonstrate improvement in the writing quality over the course of the tutorial series, which are partially supported with the findings from corpus-based analysis of participants’ texts. The findings are of relevance to e-learning scholars, faculty, and administrators of English-medium universities with substantial intakes of international graduate students in Social Sciences and Education whose native languages are other than English.

Middle Permian granitoids in the western section of the northern Qaidam Block, NW China: Petrogenesis and tectonic implications

Origin and tectonic setting of late Late Paleozoic to Early Mesozoic igneous rocks in the northern Qaidam Block (NQ) have long been a matter of debate. In this paper, we report new zircon U-Pb ages and Hf isotopic compositions and bulk rock major and trace elemental compositions for granitoids of the Yanchangbeishan area in the NQ. The granodiorite, tonalite, granodiorite, and syenogranite yielded zircon U-Pb ages of 271.9 Ma, 262.0 Ma, 272.4 Ma and 261.3 Ma, respectively, corresponding to εHf(t) values are 13.04–14.32, 8.14–10.77, 8.66–11.88, and 8.22–10.43, respectively. These granitoids are compositionally I-type granites and exhibit varying degrees of enrichment in Rb, Th, U, and K, and depletion in Nb, Ta, and Ti. According to field relationships, petrology, geochronology, and geochemistry, the granitoid types can be divided into three groups: (1) ca. 272 Ma granodiorites, which are formed through partial melting of metagreywackes and amphibolites derived from depleted mantle; (2) ca. 262 Ma tonalite, which represents a typical high-silica adakite derived from partial melting of subducted oceanic crust during the evolution of depleted mantle; and (3) ca. 261 Ma syenogranite, which is formed by the partial melting and high-degree fractional crystallization of mixed crustal materials. Considering our findings and the regional geological background, we propose a five-stage tectonic evolution model of subduction slab roll-back and subduction zone retreat for the late Late Paleozoic to Early Mesozoic of NQ.

Zircon U–Pb–Hf isotope and geochemical constraints on the petrogenesis and tectonic setting of Mesoarchean granitoids from the Carajás province, Amazonian craton, Brazil

Based on U–Pb–Hf isotope and whole-rock geochemical data, this paper investigates the origin and tectonic setting of Mesoarchean granitoids from the Ourilândia do Norte area, Carajás province (northern Brazil) and compares them with other Archean units. The following units were selected: (i) tonalite-trondhjemite-granodiorite (TTG) from the Mogno and Rio Verde suites, represented by a 2.92 Ga tonalitic xenolith [εHf(2.92 Ga) = +2.0 to –0.2] and a 2.92 Ga porphyritic trondhjemite [εHf(2.92 Ga) = +2.3 to –3.5], respectively; (ii) the Ourilândia sanukitoid suite, composed of the 2.92 Ga Arraias granodiorite [εHf(2.92 Ga) = +1.9 to –4.4] and 2.88 Ga Ourilândia tonalite-granodiorite complex [εHf(2.88 Ga) = +3.2 to –1.0]; and (iii) the Boa Sorte granite (Canaã dos Carajás granitic suite), which provides four zircon populations dated at 3.04 Ga, 2.98 Ga, 2.93 Ga and 2.88 Ga. The 2.88 Ga population represents the granitic crystallization age, with εHf(2.88 Ga) values ranging from –0.8 to –4.1. The Paleoarchean geological record from the Carajás province is scarce and restricted to detrital zircons (<3.7 Ga), rare xenocrystals (<3.7 Ga) and crustal model ages (Hf-TDMC = 3.6–3.1 Ga), which indicate Paleoarchean crust extraction from the mantle. The zircon U–Pb–Hf isotope and whole-rock geochemistry data from the Mesoarchean granitoids support the hypothesis that the Carajás province experienced mantle enrichment and progressive crustal reworking during 3.1–2.8 Ga, similar to the Mesoarchean record from the Pilbara and Kaapvaal cratons. A proposed three-stage tectonic model explains the petrogenesis and Hf isotope signatures of the studied granitoids. Regardless of the extent of the Paleoarchean geological record, the available data allow us to postulate that the formation of Paleoarchean crust in the Carajás province involved a long-lived dome-and-keel setting (3.6–3.1 Ga), and that this crust was later recycled, allowing mantle enrichment during Mesoarchean low-angle subduction (3.07–2.92 Ga). Then, a short-lived collision defined by peak regional metamorphism (2.89–2.84 Ga) and associated with crustal thickening and slab breakoff gave rise to mantle- and crust-derived magmas at 2.88 Ga.

Context learning from a ship trajectory cluster for anomaly detection

This paper presents a context information extraction process over Automatic Identification System (AIS) real-world ship data, building a system with the capability to extract representative points of a trajectory cluster. With the trajectory cluster, the study proposes the use of trajectory segmentation algorithms to extract representative points of each trajectory and then use the k-means algorithm to obtain a series of centroids over all the representative points. These centroids, combined, form a new representative trajectory of the cluster. This new representative trajectory of the input cluster represents new contextual information extracted from the original set of trajectories, being possible to apply anomaly detection approaches over the new obtained context. The results show a suitable approach with several compression algorithms that are compared with a metric based on the Perpendicular Euclidean Distance.

Robustness-Eva-MRC: Assessing and analyzing the robustness of neural models in extractive machine reading comprehension

Deep neural networks, despite their remarkable success in various language understanding tasks, have been found vulnerable to adversarial attacks and subtle input perturbations, revealing a robustness shortfall. To explore this, this paper presents Robustness-Eva-MRC, an interactive platform designed to assess and analyze the robustness of pre-trained and large-scale language models in extractive machine reading comprehension (MRC) tasks. The platform integrates eight adversarial attack methods across character-, word-, and sentence-levels, and applies them to five MRC datasets, thereby fabricating challenging adversarial testing sets. Then it evaluates the MRC models on both original and adversarial sets, yielding insights into their robustness through performance gaps. Moreover, Robustness-Eva-MRC provides comprehensive visualizations and detailed case studies, enhancing the understanding of model robustness. A screencast video and additional material are available at https://github.com/distantJing/Robustness-Eva-MRC.

FontFusionGAN: Refinement of Handwritten Fonts by Font Fusion

Handwritten fonts possess unique expressive qualities; however, their clarity often suffers because of inconsistent handwriting. This study introduces FontFusionGAN (FFGAN), a novel method that enhances handwritten fonts by mixing them with printed fonts. The proposed approach leverages a generative adversarial network (GAN) to synthesize fonts that combine the desirable features of both handwritten and printed font styles. Training a GAN on a comprehensive dataset of handwritten and printed fonts enables it to produce legible and visually appealing font samples. The methodology was applied to a dataset of handwriting fonts, showing substantial enhancements in the legibility of the original fonts, while retaining their unique aesthetic essence. Unlike the original GAN setting where a single noise vector is used to generate a sample image, we randomly selected two noise vectors, z1 and z2, from a Gaussian distribution to train the generator. Simultaneously, we input a real image into the fusion encoder for exact reconstruction. This technique ensured the learning of style mixing during training. During inference, we provided the encoder with two font images, one handwritten and the other printed font, to obtain their respective latent vectors. Subsequently, the latent vector of the handwritten font image was injected into the first five layers of the generator, whereas the latent vector of the printed font image was injected into the last two layers to obtain a refined handwritten font image. The proposed method has the potential to improve the readability of handwritten fonts, offering benefits across diverse applications, such as document composition, letter writing, and assisting individuals with reading and writing difficulties.

AffinityVAE: A multi-objective model for protein-ligand affinity prediction and drug design

In the prediction of protein-ligand affinity, the traditional methods require a large amount of computing resources, and have certain limitations in predicting and simulating the structural changes. Although employing data-driven approaches can yield favorable outcomes in deep learning, it entails a lack of interpretability. Some methods may require additional structural information or domain knowledge to support the interpretation, which may limit their applicability. This paper proposes an affinity variational autoencoder (AffinityVAE) using interaction feature mapping and a variational autoencoder, which consists of a multi-objective model capable of end-to-end affinity prediction and drug discovery. In this study, the limitations of affinity prediction in terms of interpretability are tackled by proposing the concept of a protein-ligand interaction feature map. This increases the diversity and quantity of protein-ligand binding data by designing an adaptive autoencoder of target chemical properties to generate new ligands similar to known ligands and adding them to the original training set. AffinityVAE is then retrained using this extended training set to further validate the protein-ligand binding affinity prediction. Comparisons were conducted between the AffinityVAE and recent methods to demonstrate the high efficiency of the proposed model. The experimental results show that AffinityVAE has very high prediction performance, and it has the potential to enhance the diversity and the amount of protein-ligand binding data, which promotes the drug development.

A multistart tabu search-based method for feature selection in medical applications

In the design of classification models, irrelevant or noisy features are often generated. In some cases, there may even be negative interactions among features. These weaknesses can degrade the performance of the models. Feature selection is a task that searches for a small subset of relevant features from the original set that generate the most efficient models possible. In addition to improving the efficiency of the models, feature selection confers other advantages, such as greater ease in the generation of the necessary data as well as clearer and more interpretable models. In the case of medical applications, feature selection may help to distinguish which characteristics, habits, and factors have the greatest impact on the onset of diseases. However, feature selection is a complex task due to the large number of possible solutions. In the last few years, methods based on different metaheuristic strategies, mainly evolutionary algorithms, have been proposed. The motivation of this work is to develop a method that outperforms previous methods, with the benefits that this implies especially in the medical field. More precisely, the present study proposes a simple method based on tabu search and multistart techniques. The proposed method was analyzed and compared to other methods by testing their performance on several medical databases. Specifically, eight databases belong to the well-known repository of the University of California in Irvine and one of our own design were used. In these computational tests, the proposed method outperformed other recent methods as gauged by various metrics and classifiers. The analyses were accompanied by statistical tests, the results of which showed that the superiority of our method is significant and therefore strengthened these conclusions. In short, the contribution of this work is the development of a method that, on the one hand, is based on different strategies than those used in recent methods, and on the other hand, improves the performance of these methods.

Did the Troodos ophiolite of Cyprus form during subduction inception?

Understanding whether ophiolites formed during subduction inception or above long-lived subduction zones poses an outstanding challenge as modern examples of incipient subduction zones are rare. It is generally agreed that the Troodos ophiolite of Cyprus was formed at a spreading center above a subduction zone within the closing Neo-Tethys Ocean during the Late Cretaceous; beyond this, however, details of its geodynamic origin have been widely debated. To explore this problem, we examine further the analogy previously made between the Troodos ophiolite and the Mariana fore-arc, by here comparing extrusive rocks from the Troodos ophiolite to Pliocene-aged basalts from the southeastern Mariana fore-arc rifts (SEMFR), which formed by near-trench spreading above a highly dehydrating, mature subducted slab. The SEMFR basaltic glasses and olivine-hosted melt inclusions possess the highest markers of water-rich, slab-fluids (Rb/Th, Cs/Th, H2O/Ce, Ba/Th) of any worldwide arc magmas, a signature that may be consistent with efficient slab dehydration within ∼90–100 km from the trench. Troodos ophiolite lavas, which have preserved fresh glasses, recorded similar, elevated slab-fluid markers to those of the SEMFR glasses. Although geodynamic comparisons cannot be direct, we nevertheless propose that the chemical similarities in these slab-fluid markers support the formation of the Troodos ophiolite above a highly dehydrating, shallow subducted plate in a near-trench setting. Unlike the southern Marianas, lack of juvenile arc volcanism and associated volcaniclastics, but presence of an extensive boninitic sheeted dyke complex with associated extrusive section, demonstrates that Troodos presumably developed by organized seafloor spreading in a near-trench setting, immediately following subduction inception of the Neo-Tethyan plate. As such, the chemical similarities between the SEMFR magmas and the Troodos ophiolite can only be used to provide constraints upon the geodynamic and the petrogenetic conditions of magma formation. Hence, deciphering precisely the original setting of ophiolites requires us to integrate thoroughly accurate geology and petrographic descriptions, along with comparison with modern analogues of infant subduction zones.

Ibn Rushd in the Safavid Iran: “En Orient, après Averroès … ” Revisited

Abstract Tracing the possible ways in which the thought of Ibn Rushd (Averroes, d. 595/1198) was transmitted within Muslim contexts can be such a “frustrating” and “depressing” task that it may have contributed to the near absence of his name in the lists of influential philosophers of Islam in the eastern part of the Islamic world. This article, through the consultation of original texts and manuscripts sets out to throw some light on this seeming absence. It offers a revision of Henry Corbin’s famous assertion about Ibn Rushd’s fate in the East by addressing the various receptions of Ibn Rushd’s texts in North Africa, Egypt, and the Levant, examining available manuscript catalogs of Ibn Rushd’s writings in Iran, and analyzing an important work by ʿAbd al-Razzāq al-Lāhījī (d. 1072/1661–1662), one of the most famous philosophers of the 17th century.

Evasion Generative Adversarial Network for Low Data Regimes

A myriad of recent literary works has leveraged generative adversarial networks (GANs) to generate unseen evasion samples. The purpose is to annex the generated data with the original train set for adversarial training to improve the detection performance of machine learning (ML) classifiers. The quality of generated adversarial samples relies on the adequacy of training data samples. However, in low data regimes like medical diagnostic imaging and cybersecurity, the anomaly samples are scarce in number. This paper proposes a novel GAN design called Evasion Generative Adversarial Network (EVAGAN) that is more suitable for low data regime problems that use oversampling for detection improvement of ML classifiers. EVAGAN not only can generate evasion samples, but its discriminator can act as an evasion-aware classifier. We have considered Auxiliary Classifier GAN (ACGAN) as a benchmark to evaluate the performance of EVAGAN on cybersecurity (ISCX-2014, CIC-2017 and CIC2018) botnet and computer vision (MNIST) datasets. We demonstrate that EVAGAN outperforms ACGAN for unbalanced datasets with respect to detection performance, training stability and time complexity. EVAGAN&#x0027;s generator quickly learns to generate the low sample class and hardens its discriminator simultaneously. In contrast to ML classifiers that require security hardening after being adversarially trained by GAN-generated data, EVAGAN renders it needless. The experimental analysis proves that EVAGAN is an efficient evasion hardened model for low data regimes for the selected cybersecurity and computer vision datasets. Code will be available at HTTPS://www.github.com/rhr407/EVAGAN. <i>ImpStatement</i> Artificial Intelligence (AI) applications can help improve the quality of human life. The use of AI is not only limited to medical anomaly detection and drug discovery but can be leveraged in computer networks to keep people safe from malicious activities on the Internet. However, the AI-based models can be biased towards the majority class of data on which they are trained due to data imbalance. Anomaly data samples are always scarce as compared to normal data samples. So this is an open research problem to solve. Our work is an effort to improve the AI-based methods in detection performance, stability and time complexity. Using the proposed technique, we can train our AI model using fewer anomaly samples, improving the cost-efficiency compared to state-of-the-art in anomaly detection.