Bimodal Architecture Research Articles

Multimodal learning for temporal relation extraction in clinical texts.

This study focuses on refining temporal relation extraction within medical documents by introducing an innovative bimodal architecture. The overarching goal is to enhance our understanding of narrative processes in the medical domain, particularly through the analysis of extensive reports and notes concerning patient experiences. Our approach involves the development of a bimodal architecture that seamlessly integrates information from both text documents and knowledge graphs. This integration serves to infuse common knowledge about events into the temporal relation extraction process. Rigorous testing was conducted on diverse clinical datasets, emulating real-world scenarios where the extraction of temporal relationships is paramount. The performance of our proposed bimodal architecture was thoroughly evaluated across multiple clinical datasets. Comparative analyses demonstrated its superiority over existing methods reliant solely on textual information for temporal relation extraction. Notably, the model showcased its effectiveness even in scenarios where not provided with additional information. The amalgamation of textual data and knowledge graph information in our bimodal architecture signifies a notable advancement in the field of temporal relation extraction. This approach addresses the critical need for a more profound understanding of narrative processes in medical contexts. In conclusion, our study introduces a pioneering bimodal architecture that harnesses the synergy of text and knowledge graph data, exhibiting superior performance in temporal relation extraction from medical documents. This advancement holds significant promise for improving the comprehension of patients' healthcare journeys and enhancing the overall effectiveness of extracting temporal relationships in complex medical narratives.

Versatile Technique to Produce a Hierarchical Design in Nanoporous Gold.

The potential to generate variable pore sizes, simplistic surface modification, and a breadth of commercial uses in the fields of biosensors, actuators, drug loading and release, and the development of catalysts have unquestionably accelerated the usage of nanoporous gold (NPG)-based nanomaterials in research and development. This article describes the process of the generation of hierarchical bimodal nanoporous gold (hb-NPG) by employing a step-wise procedure involving electrochemical alloying, chemical dealloying techniques, and annealing to create both macro- and mesopores. This is done to improve the utility of NPG by creating a bicontinuous solid/void morphology. The area available for surface modification is enhanced by smaller pores, while molecular transport benefits from the network of larger pores. The bimodal architecture, which is the result of a series of fabrication steps, is visualized using scanning electron microscopy (SEM) as a network of pores that are less than 100 nm in size and connected by ligaments to larger pores that are several hundred nanometers in size. The electrochemically active surface area of the hb-NPG is assessed using cyclic voltammetry (CV), with a focus on the critical roles that both dealloying and annealing play in creating the necessary structure. The adsorption of different proteins is measured by solution depletion technique, revealing the better performance of hb-NPG in terms of protein loading. By changing the surface area to volume ratio, the created hb-NPG electrode offers tremendous potential for biosensor development. The manuscript discusses a scalable method to create hb-NPG surface structures, as they offer a large surface area for the immobilization of small molecules and improved transport pathways for faster reactions.

Bi-modal email spam detection using recurrent and convolution neural network techniques

The increasing adoption of electronic emails as a means of communication, both at the commercial, government, and individual levels, serves as an impetus for attackers to compromise communication. Consequently, numerous machine learning techniques have been developed for identifying unwanted emails, commonly known as spam. Despite the significant progress reported in existing literature, most studies do not integrate the detection of both textual and image-based spam. In this paper, two deep learning techniques that detect both textual and image-based spam were evaluated. First, the Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN) is studied, training them on various textimage features to explore their effectiveness on an improved dataset. Subsequently, in an effort to outsmart current spam detection techniques, a bi-modal architecture capable of detecting textual spam, image spam, and mixed spam is designed. The experimental results in conjunction with existing transfer learning for effective spam detection is provided.

Learning English with Peppa Pig

Abstract Recent computational models of the acquisition of spoken language via grounding in perception exploit associations between spoken and visual modalities and learn to represent speech and visual data in a joint vector space. A major unresolved issue from the point of ecological validity is the training data, typically consisting of images or videos paired with spoken descriptions of what is depicted. Such a setup guarantees an unrealistically strong correlation between speech and the visual data. In the real world the coupling between the linguistic and the visual modality is loose, and often confounded by correlations with non-semantic aspects of the speech signal. Here we address this shortcoming by using a dataset based on the children’s cartoon Peppa Pig. We train a simple bi-modal architecture on the portion of the data consisting of dialog between characters, and evaluate on segments containing descriptive narrations. Despite the weak and confounded signal in this training data, our model succeeds at learning aspects of the visual semantics of spoken language.

DeePROG: Deep Attention-Based Model for Diseased Gene Prognosis by Fusing Multi-Omics Data.

An in-depth exploration of gene prognosis using different methodologies aids in understanding various biological regulations of genes in disease pathobiology and molecular functions. Interpreting gene functions at biological and molecular levels remains a daunting yet crucial task in domains such as drug design, personalized medicine, and next-generation diagnostics. Recent advancements in omics technologies have produced diverse heterogeneous genomic datasets like micro-array gene expression, miRNA expression, DNA sequence, 3D structures, which are significant resources for understanding the gene functions. In this paper, we propose a novel self-attention based deep multi-modal model, named DeePROG, for the prognosis of disease affected genes based on heterogeneous omics data. We use three NCBI datasets covering three modalities, namely gene expression profile, the underlying DNA sequence, and the 3D protein structures. To extract useful features from each modality, we develop several context-specific deep learning models. Besides, we develop three attention-based deep bi-modal architectures along with DeePROG to leverage the prognosis of the underlying biomedical data. We assess the performance of the models' in terms of computational assessment of function annotation (CAFA2) metrics. Moreover, we analyze the results in terms of receiver operating characteristics (ROC) curve in high-class imbalance data setting and perform statistical significance tests in terms of Welch's t-test. Experiment results show that DeePROG significantly outperforms baseline models across in terms of performance metrics. The source code and all preprocessed datasets used in this study are available at https://github.com/duttaprat/DeePROG.

Mechanical behavior of heterostructured iron films with precisely defined bimodal architectures

Heterostructured metals exhibit a superior combination of strength, ductility and toughness compared to both nanocrystalline (NC) and coarse-grained metals with homogeneous microstructures. However, despite considerable advances in characterizing their mechanical behavior and deformation mechanisms, only limited progress has been made in synthesizing heterostructured metals with robust microstructural control. Here, we utilize a novel technique to synthesize heterostructured Fe films with precisely defined bimodal architectures wherein the NC (grain size ∼ 50 nm) and single crystal (SC) domains (> 1000 nm) are explicitly configured to be in parallel, series and wave-like architecture. As anticipated, the films with parallel architecture exhibited the highest strain to failure due to the co-deformation of NC and SC domains. But contrary to expectation, the wave-like architecture led to lower strength and strain to failure compared to the series architecture. This discrepancy arises due to the formation of larger dislocation pile-ups in the SC domains of films with wave-like architecture, which increases the local stress concentration and leads to early yielding and premature failure. These results demonstrate that the interplay between the geometry of dislocation slip and the spatial orientation of SC (coarse) domains is a key factor in determining the mechanical behavior of heterostructured metals. This synthesis method provides a route to systematically probe the relationship between the microstructural architecture and mechanical behavior of heterostructured metals, and tailor their mechanical properties in a reproducible manner.

ФІНАНСОВІ ТЕХНОЛОГІЇ: ГЛОБАЛЬНІ ТРЕНДИ ТА УРОКИ ДЛЯ УКРАЇНИ

Trends in the penetration of financial innovations into global markets have been analyzed in this article. The prerequisites for their growth in a global format were identified when the positions of different countries were compared in terms of the Digital Adoption Index 2019. The penetration of financial innovations into global markets is proceeding at a rather fast pace as a result of the fact that some markets become less attractive or too expensive for financial institutions, especially in post-crisis conditions. The current trends point to the progressive development of the FinTech market. In the article, the opinions of various authors who conducted research in the field of FinTech were studied in order to find out the main features and characteristics that make up the essence of this term. The prerequisites for the introduction of FinTech were justified and the advantages and disadvantages for it were identified. It was found that banks and companies, when testing modern financial technologies, face the challenge limiting their actions. Therefore, attention has focused on the implementation of regulatory Sandbox. The benefits of implementing such a tool from the point of view of business, customers and government have been analyzed and summarized. Investment activity trends were studied from the point of view of investing into financial technologies in the global market in order to substantiate the prospects for the development of FinTech in the future. An analysis of the FinTech market in Ukraine was carried out to find out how this market has developed. The rapid changes that are taking place in connection with digitalization have created a field for the development of the concept of a bimodal architecture of corporate IT, but now there is a question about the imperfection of bimodal systems. Discussion of this issue is now very relevant and causes a lot of discussion, because companies need to develop a digital strategy. It is proposed to develop the parameters of the digital strategy in terms of substantiating the digital core, digital products and assets, creating communication channels and business networks.

Thermodynamic and Kinetic Transitions of Liquids in Nanoconfinement.

Core principles of chemistry are ubiquitously invoked to shed light on the nature of molecular level interactions in nanoconfined fluids, which play a pivotal role in a wide range of processes in geochemistry, biology, and engineering. A detailed understanding of the physicochemical processes involved in the flow, structural transitions, and freezing or melting behavior of fluids confined within nanometer-sized pores of solid materials is thus of enormous importance for both basic research and technological applications.This Account provides a perspective on new insights into the thermodynamic and kinetic transitions of nanoconfined fluids in their stable and metastable forms. After briefly introducing the unique properties of mesoporous silicas from the SBA, MCM, and FDU families that serve as the confinement matrices, combining highly ordered single and bimodal mesopore architectures with tunable pore sizes in the ∼2-15 nm range and narrow size distributions, recent studies on melting/freezing behavior of water confined in these host matrices are reviewed. While differential scanning calorimetry (DSC) reveals a linear relationship between melting point depression and pore size (independent of the pore shape), as predicted by the Gibbs-Thomson relation, variable temperature 2H wide-line nuclear magnetic resonance (NMR) spectroscopy studies confirm the core-shell model of water and give evidence for a layer-by-layer freezing mechanism, which gives rise to an apparent fragile-to-strong transition in the solidification dynamics.In contrast to the freezing/melting behavior of water, the effect of nanoconfinement on the glass transition of supercooled liquids is nonuniversal and the glass transition temperature Tg can either increase or decrease with the dimensionality and extent of confinement. This nonuniversal behavior is exemplified by the two glass-forming molecular liquids, glycerol and ortho-terphenyl (OTP). While glycerol shows an increase in Tg and a pronounced slowdown of the rotational dynamics of the constituent molecules due to a change in the molecular packing between the bulk and the confined liquid, OTP displays a linear and confining-media-dependent depression of Tg with increased confinement that is strongly influenced by the pore-liquid interface characteristics.This Account concludes with a focus on recent experimental evidence of extreme spatial and dynamical heterogeneity in both freezing and glass transition processes. This discovery was enabled by the unique mesoporous structures of SBA-16 and FDU-5, possessing bimodal architectures with two interconnected pore types of different size and shape (spherical and cylindrical). For the very first time, two melting points for water and two glass transitions for supercooled OTP, corresponding to a specific pore type, were observed. Collectively, these observations strongly suggest a close mechanistic connection between the local fluctuations in the structure and dynamics of nanoconfined liquids. While the findings reviewed in this Account provide new insights into thermodynamic and kinetic transitions of fluids, there remain many unanswered questions regarding the effects of nanoconfinement on the fundamental properties of fluids, which offer exciting future opportunities in chemical research.

Tunable Assembly of Block Copolymer Tethered Particle Brushes by Surface-Initiated Atom Transfer Radical Polymerization.

A strategy to synthesize SiO2-g-PMMA/PMMA-b-PS mono- and bimodal block copolymer particle brushes by surface-initiated atom transfer radical polymerization (SI-ATRP) from silica particles is presented. First, PMMA blocks were prepared by normal ATRP with controlled degree of polymerizations and grafting density. In a second step, the PS block was synthesized through a chain extension using low parts per million of Cu catalyst. Variation of the SiO2-g-PMMA-Br macroinitiator concentration had a pronounced effect on the modality of the chain extension product. In the limit of small concentration, partial termination resulted in bimodal brush architectures, while more uniform brush architectures were observed with increasing concentration of macroinitiator. Brush nanoparticles with bimodal architectures assembled into string-like aggregates that bore a resemblance to structures found in systems comprised of sparse (homopolymer) brush particles. The unexpected effect of modality on structure formation points to opportunities in controlling microstructures in brush particle materials.

Bimodal architecture and rheological and foaming properties for gamma-irradiated long-chain branched polylactides

A series of long-chain branched polylactides (LCB-PLAs) were prepared from a linear PLA precursor by applying gamma radiation with addition of the trifunctional monomer, trimethylolpropane triacrylate (TMPTA) and their topological structures were investigated by size-exclusion chromatography coupled with a multiangle light scattering detector (SEC-MALLS) and rheology. SEC-MALLS measurements show that LCB-PLAs exhibit not only the increased weight-average molecular mass but also a bimodal architecture with a short linear chain fraction and a LCB fraction, as compared with LCB-PLA samples generated by electron-beam radiation, which only show the monomodal macromolecular structure. The introduction of LCB structure contributes to the enhancement of complex viscosity, shear-thinning and storage modulus, and the deviations of phase angle, δ from the “universal” curve of linear PLA in δ(|G*|) plots (van Gurp–Palmen plot). By the analysis of the thermorheological behaviors and determination of activation energies, the bimodal architecture is confirmed. Activation energies around 77 kJ mol−1 lead to an expectation that some linear macromolecules are still present in LCB-PLAs, while the higher activation energies at large phase angles are associated with the existence of the LCB ones. A conclusion with respect to the tree-like topography for LCB-PLAs is drawn from the molecular mass dependences of zero-shear viscosity (η0–Mw plot). An explanation to these findings is provided under the consideration of the radiation dose rate for the gamma radiation. The remarkable modification of PLA topological chain structure contributes to the improvement of the foaming properties of LCB-PLAs.

Hierarchical porous gold electrodes: Preparation, characterization, and electrochemical behavior

Hierarchical porous gold films with a well-defined bimodal architecture have been made by electrodepositing gold at a constant current around a close-packed assembly of raspberry-like latex spheres (1200/60nm) followed by template removal. Electrodeposition was stopped when the gold was either ½ layer or 1½ layer thick as evident from oscillations in the potential vs time traces. Scanning electron microscopy (SEM) images show the hierarchical pore structure with an ensemble of small ∼20nm openings located in a large ∼1200nm diameter macropore. Prior to electrochemical characterization, the electrodes were cleaned either chemically and/or via UV radiation and X-ray photoelectron spectroscopy (XPS) was used to evaluate the presence of residual polystyrene. Of the three cleaning methods investigated, sonication in chloroform–acetone followed by UV radiation proved best. The surface area of the hierarchical porous gold electrodes, determined by integrating the area under the gold oxide peak, was 4× larger than a bare gold electrode and 2× larger than a macroporous gold electrode prepared using unimodal, 1200nm diameter latex spheres as the template. The electrochemical performance of the electrodes relative to the macroporous gold and flat gold was undertaken using cyclic voltammetry. The results show that the non-Faradaic current scales linearly with electrode area while the Faradaic current of a diffusing electrochemically reversible redox probe (ferrocene methanol) does not. For an adsorbed redox couple (ferrocene hexanethiol), the voltammetric wave shapes and surface coverage were different for the different electrodes.

Coagent-induced transformations of polypropylene microstructure: Evolution of bimodal architectures and cross-linked nano-particles

Polypropylene is transformed by simultaneous, radical-mediated chain scission and cross-linking to generate branched architectures. While macroradical fragmentation reduces the molar mass of the dominant chain population, cross-linking by triallyl trimesate (TAM) activation yields a minority population of hyper-branched chains that is less susceptible to molecular weight loss. This disparity in chain reactivity leads to bimodal molecular weight and branching distributions. Furthermore, a precipitation polymerization of TAM can proceed concurrently with PP branching to produce a low yield of cross-linked, TAM-rich nano-particles. The mechanisms through which unimodal composition and molecular weight distributions evolve toward a bimodal condition are discussed, along with the factors that lead to particle formation.

RBS: A ROBUST BIMODAL SYSTEM FOR FACE RECOGNITION

During the last few years, many algorithms have been proposed in particular for face recognition using classical 2-D images. However, it is necessary to deal with occlusions when the subject is wearing sunglasses, scarves and such. In the same way, ear recognition is arising as a new promising biometric for people recognition, even if the related literature appears to be somewhat underdeveloped. In this paper, several hybrid face/ear recognition systems are investigated. The system is based on IFS (Iterated Function Systems) theory that are applied on both face and ear resulting in a bimodal architecture. One advantage is that the information used for the indexing and recognition task of face/ear can be made local, and this makes the method more robust to possible occlusions. The distribution of similarities in the input images is exploited as a signature for the identity of the subject. The amount of information provided by each component of the face and the ear image has been assessed, first independently and then jointly. At last, results underline that the system significantly outperforms the existing approaches in the state of the art.