Training data augmentation: An empirical study using generative adversarial net-based approach with normalizing flow models for materials informatics

Biological systems and processes have had, and continue to have, important implications and applications in materials extraction, processing, and performance. This paper illustrates some interdisciplinary, biological issues in materials science and engineering. These include metal extraction involving bacterial catalysis, galvanic couples, bacterial-assisted corrosion and degradation of materials, biosorption and bioremediation of toxic and other heavy metals, metal and material implants and prostheses and related dental and medical biomaterials developments and applications, nanomaterials health benefits and toxicity issue, and biomimetics and biologically inspired materials developments. These and other examples provide compelling evidence and arguments for emphasizing biological sicences in materials science and engineering curricula and the implementation of a bio-materials paradigm to facilitate the emergence of innovative interdisciplinarity involving the biological sciences and materials sciences and engineering.

High-density crowd events like public concerts, sporting events, or religious festivals represent significant safety challenges due to high crowd density. Methods of monitoring crowds using manual observation or passive surveillance often don't provide real-time information. Therefore, we present a real-time crowd density estimation solution that utilizes YOLOv8 for people detection and CSRNet for density estimation. The crowd density estimation system utilizes live video feeds from direct surveillance cameras/CCTV or drone footage. The system will assess the crowd density in four levels across four distinct areas: Low, Medium, High, or Critical crowd density. The density estimator has a web-based dashboard that provides real-time analytics reports, heatmap density estimates, and historical records, which can assist in making quick, informed decisions and assessments following an event. The system is validated on benchmark datasets and real-world video streams with 95.3% detection accuracy, 7.4 MAE in crowd counting and 28 FPS processing with off-the-shelf GPU hardware. The results show high accuracy with low latency, making it feasible for real-world applications for large-scale events. The main contributions of the work include using YOLOv8 integrated with CSRNet to jointly detect and estimate crowds, developing a real-time dashboard to provide transparent crowd analytics, and system validation with quantitative metrics and real-world evidence.

As a new way of training generative models, Generative Adversarial Net (GAN) that uses a discriminative model to guide the training of the generative model has enjoyed considerable success in generating real-valued data. However, it has limitations when the goal is for generating sequences of discrete tokens. A major reason lies in that the discrete outputs from the generative model make it difficult to pass the gradient update from the discriminative model to the generative model. Also, the discriminative model can only assess a complete sequence, while for a partially generated sequence, it is non-trivial to balance its current score and the future one once the entire sequence has been generated. In this paper, we propose a sequence generation framework, called SeqGAN, to solve the problems. Modeling the data generator as a stochastic policy in reinforcement learning (RL), SeqGAN bypasses the generator differentiation problem by directly performing gradient policy update. The RL reward signal comes from the GAN discriminator judged on a complete sequence, and is passed back to the intermediate state-action steps using Monte Carlo search. Extensive experiments on synthetic data and real-world tasks demonstrate significant improvements over strong baselines.

Collaborative filtering based recommendation learns users’ preferences from all users’ historical behavior data, and has been popular to facilitate decision making. Recently, the fairness issue of recommendation has become more and more essential. A recommender system is considered unfair when it does not perform equally well for different user groups according to users’ sensitive attributes (e.g., gender, race). Plenty of methods have been proposed to alleviate unfairness by optimizing a predefined fairness goal or changing the distribution of unbalanced training data. However, they either suffered from the specific fairness optimization metrics or relied on redesigning the current recommendation architecture. In this paper, we study how to improve recommendation fairness from the data augmentation perspective. The recommendation model amplifies the inherent unfairness of imbalanced training data. We augment imbalanced training data towards balanced data distribution to improve fairness. Given each real original user-item interaction record, we propose the following hypotheses for augmenting the training data: each user in one group has a similar item preference (click or non-click) as the item preference of any user in the remaining group. With these hypotheses, we generate “fake" interaction behaviors to complement the original training data. After that, we design a bi-level optimization target, with the inner optimization generates better fake data to augment training data with our hypotheses, and the outer one updates the recommendation model parameters based on the augmented training data. The proposed framework is generally applicable to any embedding-based recommendation, and does not need to pre-define a fairness metric. Extensive experiments on two real-world datasets clearly demonstrate the superiority of our proposed framework. We publish the source code at https://github.com/newlei/FDA.

Outside the highly publicized victories in the game of Go, there have been numerous successful applications of deep learning in the fields of information retrieval, computer vision, and speech recognition. In cybersecurity, an increasing number of companies have begun exploring the use of deep learning (DL) in a variety of security tasks with malware detection among the more popular. These companies claim that deep neural networks (DNNs) could help turn the tide in the war against malware infection. However, DNNs are vulnerable to adversarial samples, a shortcoming that plagues most, if not all, statistical and machine learning models. Recent research has demonstrated that those with malicious intent can easily circumvent deep learning-powered malware detection by exploiting this weakness. To address this problem, previous work developed defense mechanisms that are based on augmenting training data or enhancing model complexity. However, after analyzing DNN susceptibility to adversarial samples, we discover that the current defense mechanisms are limited and, more importantly, cannot provide theoretical guarantees of robustness against adversarial sampled-based attacks. As such, we propose a new adversary resistant technique that obstructs attackers from constructing impactful adversarial samples by randomly nullifying features within data vectors. Our proposed technique is evaluated on a real world dataset with 14,679 malware variants and 17,399 benign programs. We theoretically validate the robustness of our technique, and empirically show that our technique significantly boosts DNN robustness to adversarial samples while maintaining high accuracy in classification. To demonstrate the general applicability of our proposed method, we also conduct experiments using the MNIST and CIFAR-10 datasets, widely used in image recognition research.

Development of hydrogels with excellent and tunable mechanical properties combining with multifunctions is an intriguing issue in material science and engineering. Herein, bioinspired tunable sacrificial bonds are introduced into the tetra‐poly(ethylene glycol) (PEG) based polyurethane (PU) (TP) network to afford a hydrogel with tunable mechanical properties, shape‐memory, and self‐healing functions. The mussel‐inspired compound of Lysine‐dopamine (LDA) is introduced into the network of TP hydrogel through polyurethane/polyurea chemistry to form LDA‐tetra‐PEG‐PU (LTP) hydrogel. As catechol groups in LDAs can intermolecularly interact with each other and can also coordinate with ferric ions with different coordination ratios, these physical interactions with different strengths in the afforded LTP hydrogel construct kinds of sequentially tuned sacrificial bonds. As a result, these sacrificial bonds preferentially rupture prior to the covalent network upon external loading, which dissipate the energy and endow the hydrogel with advanced and postadjustable mechanical properties. This mechanism is investigated in detail. Furthermore, the LTP hydrogel shows multifunctions such as shape‐memory and self‐healing abilities. In addition, the tetra‐PEG based hydrogel shows remarkable thermoresponsiveness that the hydrogel distinctly contracts with the increase of the temperature. The improved mechanical strength and multifunctions should enlarge the application areas of the tetra‐PEG based hydrogel in various fields.

120 Years of nickel-based cathodes for alkaline batteries

Metal–insulator (MI) transitions in correlated electron systems have long been a central and controversial issue in material science. Vanadium dioxide (VO2) exhibits a first-order MI transition at 340 K. For more than half a century, it has been debated whether electron correlation or the structural instability due to dimerised V ions is the more essential driving force behind this MI transition. Here, we show that an ultrahigh magnetic field of 500 T renders the insulator phase of tungsten (W)-doped VO2 metallic. The spin Zeeman effect on the d electrons of the V ions dissociates the dimers in the insulating phase, resulting in the delocalisation of electrons. As the Mott–Hubbard gap essentially does not depend on the spin degree of freedom, the structural instability is likely to be the more essential driving force behind the MI transition.

Today, computational models are extensively applied in scientific research to address fundamental as well as practical issues in material science. Fullerenes and metallofullerenes are well recognized as potential materials for next-generation electronic devices, such as quantum computers. The rapid advance in fullerene/metallofullerene research requires improved theoretical tools to understand fundamental as well as practical applications of these new members of fullerene/metallofullerene family. In this presentation, we will describe the extension of the famous spiral program by Fowler and Manolopoulos1 to generate new non-classical isomers. We have extended this approach beyond 5 and 6 membered fullerene structures to other polygon faces, such as, heptagons. In addition, we have developed a machine learning tool based on a general relationship between the structure patterns and physical properties. Finally, we will describe our current modeling studies regarding chromatographic retention mechanism(s) of fullerene/metallofullerene molecules. Reference: Fowler, Patrick W., and D. E. Manolopoulos. An Atlas of Fullerenes. Clarendon Press-Oxford, 1995.

A major issue in materials science is why glasses present low-temperature thermal and vibrational properties that sharply differ from those of crystals. In particular, long-wavelength phonons are considerably more damped in glasses, yet it remains unclear how structural disorder at atomic scales affects such a macroscopic phenomenon. A plausible explanation is that phonons are scattered by local elastic heterogeneities that are essentially uncorrelated in space, a scenario known as Rayleigh scattering, which predicts that the damping of acoustic phonons scales with wavenumber k as kd+1 (in dimension d). Here we demonstrate that phonon damping scales instead as - kd+1 lnk, with this logarithmic enhancement originating from long-range spatial correlations of elastic disorder caused by similar stress correlations. Our work suggests that the presence of long-range spatial correlations of local stress and elasticity may well be the crucial feature that distinguishes amorphous solids from crystals.

Prediction of fracture toughness of ceramic composites as function of microstructure: I. Numerical simulations

Unravelling dislocation networks in metals

A novel method to study recrystallization behavior: Continuous heating stress relaxation

The determination of valuable thermodynamic data for materials at elevated temperatures is a very important issue in Materials Science. This can be achieved by the elucidation of vaporization processes with high temperature mass spectrometry. Knudsen Effusion Mass Spectrometry provides the identification of gaseous species present above condensed phases and permits the determination of their partial pressures up to a total pressure of 200 Pa. The key for establishing model calculations using DG minimization tools like FactSage, Pandat or Thermocalc is the knowledge of the thermodynamic data of molecules and compounds. This includes the partial pressures of gaseous species, the standard enthalpies and entropies of formation of gaseous and condensed phases as well as their enthalpies and entropies of mixing and the knowledge of phases and chemical compounds which are formed upon reactions. The method will be described and examples and an insight will be given how basic thermodynamic data are derived with the above mentioned method for the incorporation into thermodynamic databases. Examples of investigations carried out in our lab will be given.

Knowledge of the complex deformation behavior in the anisotropic materials is one of essential issues in materials science and it is crucial for the applications of a given material. In this study, mechanical response of Fe3(wt.%)Si single crystal to nanoindentation with spherical indenter was investigated. Hardness and indentation Young´s modulus were determined experimentally and by finite element modelling. Observed pop-in phenomenon, shape of the residual imprints and origin of the slip lines were explained on the basis of resolved shear stress computed by finite element model.