Intrinsic Complexity Research Articles

Hierarchical extreme learning machine with L21-norm loss and regularization

Recently, multilayer extreme learning machine (ELM) algorithms have been extensively studied for hierarchical abstract representation learning in the ELM community. In this paper, we investigate the specific combination of $$L_{21}$$ -norm based loss function and regularization to improve the robustness and the sparsity of multilayer ELM. As we all known, the mean square error (MSE) cost function (or squared $$L_{2}$$ -norm cost function) is commonly used as optimization cost function for ELM, but it is sensitive to outliers and impulsive noises that are pervasive in real-world data. Our $$L_{21}$$ -norm loss function can lessen the harmful influence caused by noises and outliers and enhance robustness and stability of the learned model. Additionally, the row sparse inducing $$L_{21}$$ -norm regularization can learn the most-relevant sparse representation and reduce the intrinsic complexity of the learning model. We propose a specific combination of $$L_{21}$$ -norm loss function and regularization ELM auto-encoder (LR21-ELM-AE), and then stack LR21-ELM-AE hierarchically to construct the hierarchical extreme learning machine (H-LR21-ELM). Experiments conducted on several well-known benchmark datasets are presented, the results show that the proposed H-LR21-ELM can generate a more robust, more discriminative and sparser model compared with the other state-of-the-art multilayer ELM algorithms.

Preface: De-Marging Methodologies, Blurring Media-Texts

In this preface I try to briefly illustrate the methodological perspective with which this special issue of Between Journal, edited by Hans-Joachim Backe, Massimo Fusillo and myself, analyses the intrinsic complexity of the concepts of inter-mediality, trans-mediality and cross-mediality. The aim of this research is to further extend the field under investigation interlinking directions and perspectives, methods and approaches, to understanding the hermeneutic dimension of inter-trans-cross-mediality. Concentrating on the complexity of the texts and present-day medial contents, we have tried to contemplate actual methodology, gathering together contributions from different sectors scholars (comparative literature, film, Tv and media studies, game studies, etc.), with particular attention to the innovations brought about by the ubiquity of digital technology and the osmosis between cultural institutions and productive processes.

Coordination Number Regulation of Molybdenum Single-Atom Nanozyme Peroxidase-like Specificity

Summary Nanozymes are promising alternatives to natural enzymes, but their use remains limited owing to poor specificity. Overcoming this and controlling the targeted enzyme-like performance of traditional nanozymes is extremely challenging due to the intrinsic structural complexity of these systems. We report theoretical design and experimental realization of a series of heterogeneous molybdenum single-atom nanozymes (named MoSA–Nx–C), wherein we find that the peroxidase-like specificity is well regulated by the coordination numbers of single Mo sites. The resulting MoSA–N3–C catalyst shows exclusive peroxidase-like behavior. It achieves this behavior via a homolytic pathway, whereas MoSA–N2–C and MoSA–N4–C catalysts have a different heterolytic pathway. The mechanism of this coordination-number-dependent enzymatic specificity is attributed to geometrical structure differences and orientation relationships of the frontier molecular orbitals toward these MoSA–Nx–C peroxidase mimics. This study demonstrates the rational design of peroxidase-specific nanozymes and precise regulation of their enzymatic properties.

Open versus Interpenetrated: Switchable Supramolecular Trajectories in Mechanosynthesis of a Halogen-Bonded Borromean Network

Summary Precise control over topologically intricate molecular architectures remains an open challenge for chemists due to their inherent structural complexity. We report a simple solvent-free strategy to selectively prepare two multi-component supramolecular crystalline systems based on the halogen bond, endowed with either an unknot topology or a Borromean-type entanglement. Real-time in situ monitoring of the solvent-free mechanochemical synthesis of these three-component halogen-bonded ionic co-crystals reveals that the choice of milling conditions leads to a switch in the supramolecular reaction trajectory, resulting in the selective formation of an open halogen-bonded network or a halogen-bonded Borromean-type assembly.

A simple and effective predictor to design novel fluorite-structured High Entropy Oxides (HEOs)

High-Entropy Oxides (HEOs) are a totally new class of ceramic materials that have recently attracted many scientific attentions. However, the huge intrinsic complexity and the massive number of possible combinations characterizing such systems make it hard to predict a priori their properties and their crystal structures. Moreover, the idea of designing and engineering new materials by using entropy as a driving force is conceptually exciting and intellectually stimulating. Thus, we acknowledged that predicting and synthesizing unknown entropy-stabilized single-phases of a given formula in a given crystal structure could be of great interest to the HEOs research community and, through a systematic study of 18 samples of equimolar 5-component Rare Earths-based oxides, we were able to elaborate a simple and effective predictive model to design HEOs stabilized in a single-phase fluorite-like structure.The novelty of our model, other than its simplicity and immediacy, consists in pointing out that the “dispersion” of the cationic radii of the involved elements of a certain system (expressed in terms of their standard deviation) is crucial for stabilizing fluorite-structured HEOs. Definitely, for systems owning standard deviations of the involved elements cationic radii (coordination VIII) distribution higher than 0.095, single-phase fluorite-structured systems are formed; otherwise, for s < 0.095 firstly biphasic (fluorite and bixbyite) systems are formed and then single-phase bixbyite-structured systems are formed.

Measuring Complexity in Financial Data

Stock market is a canonical example of complex systems, where a large number of interacting agents lead to joint evolution of stock returns and the collective market behavior exhibits emergent properties. However, quantifying complexity in the stock market data has been a challenging task. In this report, we explore four different measures to characterize the intrinsic complexity by evaluating structural relationship among stock returns. The first two measures are based on linear and non-linear comovement structure (accounting for contemporaneous and Granger-causal relationships), the third one is based on algorithmic complexity and the fourth one is based on spectral analysis of interacting dynamical systems. In our dataset comprising daily prices of a large number of stocks in the complete historical data of NASDAQ (1972- 2018), we see that the third and fourth measures are able to identify the largest global economic downturn in 2007-09 and associated spillovers, substantially more accurately than the first two measures. Finally, we conclude this brief research report with discussions on implications of such quantification for risk management in complex systems.

Land Degradation and Mitigation Policies in the Mediterranean Region: A Brief Commentary

Land degradation is more evident where conditions of environmental vulnerability already exist because of arid climate and unsustainable forms of land exploitation. Consequently, semi-arid and dry areas have been identified as vulnerable land, requiring attention from both science and policy perspectives. In some regions, such as the Mediterranean region, land degradation is particularly intense, although there are no extreme ecological conditions. In these contexts, a wide range of formal and informal responses is necessary to face particularly complex and spatially differentiated territorial processes. However, the fit of responses has been demonstrated to be different over time and space according to the underlying socioeconomic context and the specific ecological conditions. The present commentary discusses this sort of “entropy” in the policy response to land degradation in Southern Europe, outlining the intrinsic complexity of human–nature dynamics at the base of such processes. Reflecting the need of differentiated regional strategies and more specific national measures to combat desertification, three policy frameworks (agro-environmental, economic, social) with an indirect impact on fighting land degradation have been considered, delineating the importance of policy assemblages. Finally, the importance of policy impact assessment methodologies was highlighted, focusing on the possible responses reinforcing a continental strategy against land degradation. By evidencing the role of participatory planning, developmental policies indirectly addressing land degradation reveal to be an important vector of more specific measures abating desertification risk, creating, in turn, a favorable context for direct interventions of mitigation or adaptation to climate change.

A Simple and Effective Predictor to Design Novel Fluorite-Structured High Entropy Oxides (HEOs)

High-Entropy Oxides (HEOs) are a totally new class of ceramic materials that have recently attracted many scientific attentions. However, the huge intrinsic complexity and the massive number of possible combinations characterizing such systems make it hard to predict a priori their properties and their crystal structures. Moreover, the idea of designing and engineering new materials by using entropy as a driving force is conceptually exciting and intellectually stimulating. Thus, we acknowledged that predicting and synthesizing unknown entropy-stabilized single-phases of a given formula in a given crystal structure could be of great interest to the HEOs research community and, through a systematic study of 18 samples of equimolar 5-component Rare Earths-based oxides, we were able to elaborate a simple and effective predictive model to design HEOs stabilized in a single-phase fluorite-like structure.

On the use of virtual immersive reality for discrete choice experiments to modelling pedestrian behaviour

Modelling people's behaviour is a complex task, not only because of their intrinsic complexity but also because of their interaction with the environment and other individuals. The traditional format for discrete choice experiments involves the use of text and images. However, there is a growing tendency for using tools that offer a more realistic representation of complex and dynamic attributes in discrete choice experiments. The Virtual Immersive Reality Environment (VIRE) emerges as a resource for a better understanding and presentation of variables that are difficult to understand using text-only experiments. Despite its advantages, VIRE experiments are costly and have additional complications in their practical use. This study aims to analyse the benefits of using VIRE in discrete choice modelling, comparing with the traditional text-only format with image surveys in two contexts. The first context considers the study of pedestrian behaviour when choosing alternatives for crossing an urban street. The second context deals with the behaviour of people during an emergency exit situation inside a covered sports arena. The results suggest that the assisted use of VIRE allows respondents to perceive environmental dynamics better than traditional choice experiments. VIRE adds realism and seems to improve a respondent's cognitive understanding of complex elements in the environment created by the modeller.

Gastrointestinal Stromal Tumor: Challenges and Opportunities for a New Decade.

Gastrointestinal stromal tumor (GIST) provides a paradigm to evaluate new molecularly targeted therapies and to identify structural and functional mechanisms for drug response and resistance. Drug development in GIST has successfully exploited the high reliance on KIT/PDGFRA oncogenic signaling as a therapeutic vulnerability. The recent arrival of avapritinib and ripretinib to the GIST arena has aimed to further improve on precision kinase inhibition and address tumor heterogeneity in imatinib-resistant GIST. The two main clinical challenges for the forthcoming years entail tumor eradication in patients with early-stage GIST, and maximization of tumor response in late-stage disease. To succeed, we will need to better understand the mechanisms behind adaptation to KIT inhibition and apoptosis evasion, tumor evolution after successive lines of treatment, and to explore clinically novel creative therapeutic strategies, with the overarching goal to tackle the intrinsic oncogenic complexity while minimizing adverse events.

Nanobubbles as corrosion inhibitor in acidic geothermal fluid

Metallic corrosion is a major issue that leads to an efficiency loss and eventual failure of the system in geothermal power plants. Despite the growing understanding of mechanisms of corrosion, inhibiting steel corrosion in the acidic geothermal fluids, remains to present formidable challenges due to its intrinsic physicochemical complexity. Here, we study the use of nanobubbles as a possible corrosion inhibitor by testing alteration of the low-carbon steel plates immersed in acidic geothermal water with continuously injected air-nanobubbles. Nanobubbles have been used in a broad range of areas as they are eco-friendly, low-cost, easy-to-use and high-functional materials. We, for the first time to our knowledge, found that air-nanobubbles could inhibit steel corrosion, with inhibition efficiency of up to 50 % in the studied acidic geothermal fluid. Air-nanobubbles could act as a nanoscopic coating material in the acidic geothermal fluid, through generating a bubble mattress and/or promoting nucleation and aggregation of a very small quantity of silica precipitation on the surface of steel plates. Our finding suggests that nanobubbles can inhibit steel corrosion in various chemically different geothermal fluids, highlighting the physicochemical significance of nanobubbles as the coating material for inhibiting metal degradation in the geothermal infrastructures.

Tutorial: a computational framework for the design and optimization of peripheral neural interfaces.

Peripheral neural interfaces have been successfully used in the recent past to restore sensory-motor functions in disabled subjects and for the neuromodulation of the autonomic nervous system. The optimization of these neural interfaces is crucial for ethical, clinical and economic reasons. In particular, hybrid models (HMs) constitute an effective framework to simulate direct nerve stimulation and optimize virtually every aspect of implantable electrode design: the type of electrode (for example, intrafascicular versus extrafascicular), their insertion position and the used stimulation routines. They are based on the combined use of finite element methods (to calculate the voltage distribution inside the nerve due to the electrical stimulation) and computational frameworks such as NEURON ( https://neuron.yale.edu/neuron/ ) to determine the effects of the electric field generated on the neural structures. They have already provided useful results for different applications, but the overall usability of this powerful approach is still limited by the intrinsic complexity of the procedure. Here, we illustrate a general, modular and expandable framework for the application of HMs to peripheral neural interfaces, in which the correct degree of approximation required to answer different kinds of research questions can be readily determined and implemented. The HM workflow is divided into the following tasks: identify and characterize the fiber subpopulations inside the fascicles of a given nerve section, determine different degrees of approximation for fascicular geometries, locate the fibers inside these geometries and parametrize electrode geometries and the geometry of the nerve-electrode interface. These tasks are examined in turn, and solutions to the most relevant issues regarding their implementation are described. Finally, some examples related to the simulation of common peripheral neural interfaces are provided.

Asymmetric Spirocyclization Enabled by Iridium and Brønsted Acid-Catalyzed Formal Reductive Cycloaddition

A catalytic, enantioselective spirocyclization of formanilides or formylindolines and enamides has been developed herein. The reaction proceeds through a sequential iridium-catalyzed hydrosilylatio...

Short k-mer abundance profiles yield robust machine learning features and accurate classifiers for RNA viruses.

High-throughput sequencing technologies have greatly enabled the study of genomics, transcriptomics and metagenomics. Automated annotation and classification of the vast amounts of generated sequence data has become paramount for facilitating biological sciences. Genomes of viruses can be radically different from all life, both in terms of molecular structure and primary sequence. Alignment-based and profile-based searches are commonly employed for characterization of assembled viral contigs from high-throughput sequencing data. Recent attempts have highlighted the use of machine learning models for the task, but these models rely entirely on DNA genomes and owing to the intrinsic genomic complexity of viruses, RNA viruses have gone completely overlooked. Here, we present a novel short k-mer based sequence scoring method that generates robust sequence information for training machine learning classifiers. We trained 18 classifiers for the task of distinguishing viral RNA from human transcripts. We challenged our models with very stringent testing protocols across different species and evaluated performance against BLASTn, BLASTx and HMMER3 searches. For clean sequence data retrieved from curated databases, our models display near perfect accuracy, outperforming all similar attempts previously reported. On de novo assemblies of raw RNA-Seq data from cells subjected to Ebola virus, the area under the ROC curve varied from 0.6 to 0.86 depending on the software used for assembly. Our classifier was able to properly classify the majority of the false hits generated by BLAST and HMMER3 searches on the same data. The outstanding performance metrics of our model lays the groundwork for robust machine learning methods for the automated annotation of sequence data.

Cross-frequency and iso-frequency estimation of functional corticomuscular coupling after stroke.

Functional corticomuscular coupling (FCMC) between the brain and muscles has been used for motor function assessment after stroke. Two types, iso-frequency coupling (IFC) and cross-frequency coupling (CFC), are existed in sensory-motor system for healthy people. However, in stroke, only a few studies focused on IFC between electroencephalogram (EEG) and electromyogram (EMG) signals, and no CFC studies have been found. Considering the intrinsic complexity and rhythmicity of the biological system, we first used the wavelet package transformation (WPT)to decompose the EEG and EMG signals into several subsignals with different frequency bands, and then applied transfer entropy (TE)to analyze the IFC and CFC relationship between each pair-wise subsignal. In this study, eight stroke patients and eight healthy people were enrolled. Results showed that both IFC and CFC still existed in stroke patients (EEG → EMG: 1:1, 3:2, 2:1; EMG → EEG: 1:1, 2:1, 2:3, 3:1). Compared with the stroke-unaffected side and healthy controls, the stroke-affected side yielded lower alpha, beta and gamma synchronization (IFC: beta; CFC: alpha, beta and gamma). Further analysis indicated that stroke patients yielded no significant difference of the FCMC between EEG → EMG and EMG → EEG directions. Our study indicated that alpha and beta bands were essential to concentrating and maintaining the motor capacities, and provided a new insight in understanding the propagation and function in the sensory-motor system.

Functional compounds of crude oil during low salinity water injection

Due to the intrinsic complexity of crude oil, the advanced chemical compositional analysis would be required to detect the crucial interactions that may take place between water and crude oil during low salinity water injection (LSWI). In this study, a series of analytical techniques were combined to discover the functional compounds of crude oil contributing to low salinity water effect (LSE) during LSWI. The Fourier Transform Infrared (FT-IR) analysis and Fourier Transform Ion Cyclotron Resonance Mass Spectroscopy (FT-ICR MS) were deployed to characterise the oil/water interface of a chosen crude oil with a high potency toward formation of water microdispersion. Using the negative and positive electrospray ionisation modes (−ESI and +ESI, respectively), acidic compounds with aliphatic nature and asphaltene molecules were determined to be the most functional compounds at the interface promoting the spontaneous formation of water-in-oil microdispersion. These species are key to designing any waterflood operation in oil reservoirs.

Accurate Computation of the Absorption Spectrum of Chlorophyll a with Pair Natural Orbital Coupled Cluster Methods.

The ability to accurately compute low-energy excited states of chlorophylls is critically important for understanding the vital roles they play in light harvesting, energy transfer, and photosynthetic charge separation. The challenge for quantum chemical methods arises both from the intrinsic complexity of the electronic structure problem and, in the case of biological models, from the need to account for protein–pigment interactions. In this work, we report electronic structure calculations of unprecedented accuracy for the low-energy excited states in the Q and B bands of chlorophyll a. This is achieved by using the newly developed domain-based local pair natural orbital (DLPNO) implementation of the similarity transformed equation of motion coupled cluster theory with single and double excitations (STEOM-CCSD) in combination with sufficiently large and flexible basis sets. The results of our DLPNO–STEOM-CCSD calculations are compared with more approximate approaches. The results demonstrate that, in contrast to time-dependent density functional theory, the DLPNO–STEOM-CCSD method provides a balanced performance for both absorption bands. In addition to vertical excitation energies, we have calculated the vibronic spectrum for the Q and B bands through a combination of DLPNO–STEOM-CCSD and ground-state density functional theory frequency calculations. These results serve as a basis for comparison with gas-phase experiments.

New Multiphase Chemical Processes Influencing Atmospheric Aerosols, Air Quality, and Climate in the Anthropocene.

ConspectusAtmospheric aerosols and fine particulate matter (PM2.5) are strongly affecting human health and climate in the Anthropocene, that is, in the current era of globally pervasive and rapidly increasing human influence on planet Earth. Poor air quality associated with high aerosol concentrations is among the leading health risks worldwide, causing millions of attributable excess deaths and years of life lost every year. Besides their health impact, aerosols are also influencing climate through interactions with clouds and solar radiation with an estimated negative total effective radiative forcing that may compensate about half of the positive radiative forcing of carbon dioxide but exhibits a much larger uncertainty. Heterogeneous and multiphase chemical reactions on the surface and in the bulk of solid, semisolid, and liquid aerosol particles have been recognized to influence aerosol formation and transformation and thus their environmental effects. However, atmospheric multiphase chemistry is not well understood because of its intrinsic complexity of dealing with the matter in multiple phases and the difficulties of distinguishing its effect from that of gas phase reactions.Recently, research on atmospheric multiphase chemistry received a boost from the growing interest in understanding severe haze formation of very high PM2.5 concentrations in polluted megacities and densely populated regions. State-of-the-art models suggest that the gas phase reactions, however, are not capturing the high concentrations and rapid increase of PM2.5 observed during haze events, suggesting a gap in our understanding of the chemical mechanisms of aerosol formation. These haze events are characterized by high concentrations of aerosol particles and high humidity, especially favoring multiphase chemistry. In this Account, we review recent advances that we have made, as well as current challenges and future perspectives for research on multiphase chemical processes involved in atmospheric aerosol formation and transformation. We focus on the following questions: what are the key reaction pathways leading to aerosol formation under polluted conditions, what is the relative importance of multiphase chemistry versus gas-phase chemistry, and what are the implications for the development of efficient and reliable air quality control strategies? In particular, we discuss advances and challenges related to different chemical regimes of sulfate, nitrate, and secondary organic aerosols (SOAs) under haze conditions, and we synthesize new insights into the influence of aerosol water content, aerosol pH, phase state, and nanoparticle size effects. Overall, there is increasing evidence that multiphase chemistry plays an important role in aerosol formation during haze events. In contrast to the gas phase photochemical reactions, which are self-buffered against heavy pollution, multiphase reactions have a positive feedback mechanism, where higher particle matter levels accelerate multiphase production, which further increases the aerosol concentration resulting in a series of record-breaking pollution events. We discuss perspectives to fill the gap of the current understanding of atmospheric multiphase reactions that involve multiple physical and chemical processes from bulk to nanoscale and from regional to global scales. A synthetic approach combining laboratory experiments, field measurements, instrument development, and model simulations is suggested as a roadmap to advance future research.

Mutual modulation between surface chemistry and bulk microstructure within secondary particles of nickel-rich layered oxides

Surface lattice reconstruction is commonly observed in nickel-rich layered oxide battery cathode materials, causing unsatisfactory high-voltage cycling performance. However, the interplay of the surface chemistry and the bulk microstructure remains largely unexplored due to the intrinsic structural complexity and the lack of integrated diagnostic tools for a thorough investigation at complementary length scales. Herein, by combining nano-resolution X-ray probes in both soft and hard X-ray regimes, we demonstrate correlative surface chemical mapping and bulk microstructure imaging over a single charged LiNi0.8Mn0.1Co0.1O2 (NMC811) secondary particle. We reveal that the sub-particle regions with more micro cracks are associated with more severe surface degradation. A mechanism of mutual modulation between the surface chemistry and the bulk microstructure is formulated based on our experimental observations and finite element modeling. Such a surface-to-bulk reaction coupling effect is fundamentally important for the design of the next generation battery cathode materials.

When machine learning meets multiscale modeling in chemical reactions.

Due to the intrinsic complexity and nonlinearity of chemical reactions, direct applications of traditional machine learning algorithms may face many difficulties. In this study, through two concrete examples with biological background, we illustrate how the key ideas of multiscale modeling can help to greatly reduce the computational cost of machine learning, as well as how machine learning algorithms perform model reduction automatically in a time-scale separated system. Our study highlights the necessity and effectiveness of an integration of machine learning algorithms and multiscale modeling during the study of chemical reactions.