Space Model Research Articles

Optimizing data acquisition: a Bayesian approach for efficient machine learning model training

Acquiring a substantial number of data points for training accurate machine learning (ML) models is a big challenge in scientific fields where data collection is resource-intensive. Here, we propose a novel approach for constructing a minimal yet highly informative database for training ML models in complex multi-dimensional parameter spaces. To achieve this, we mimic the underlying relation between the output and input parameters using Gaussian process regression (GPR). Using a set of known data, GPR provides predictive means and standard deviation for the unknown data. Given the predicted standard deviation by GPR, we select data points using Bayesian optimization to obtain an efficient database for training ML models. We compare the performance of ML models trained on databases obtained through this method, with databases obtained using traditional approaches. Our results demonstrate that the ML models trained on the database obtained using Bayesian optimization approach consistently outperform the other two databases, achieving high accuracy with a significantly smaller number of data points. Our work contributes to the resource-efficient collection of data in high-dimensional complex parameter spaces, to achieve high precision ML predictions.

Time-Dependent Multireference Coupled-Cluster Method (TDMRCCM) for the Bath-Dynamics in System-Bath Approach to Nonadiabatic Dynamics.

The time-dependent multireference coupled-cluster method (TDMRCCM) fits well in the scheme of the system-bath separation used to study the nonadiabatic dynamics. In TDMRCCM, a projection operator is defined as one that projects the full Hilbert space onto the space spanned by the collection of system degrees of freedom, called the model space. The inverse of this projection operator is a wave operator that acts on the model space and takes its projection back to the full wave function. This wave operator is defined as an exponential of the excitation terms and, hence, can be expanded into a Taylor series, which we have truncated in this work at the second-order of excitations. The attraction of using TDMRCCM for describing the bath dynamics is due to the exponential nature of the ansatz used in the method, which makes it possible for the higher-order excitations to be absorbed by the lower-order terms, even upon truncating the series. This improves the accuracy of the numerical calculations using TDMRCCM as an approximation for the bath-mode dynamics in the system-bath framework for nonadiabatic dynamics. We present the theoretical details of TDMRCCM and the numerical results for implementing this method to study the dynamics in the butatriene cation.

Predictive modelling of freeway space utilising clinical history, normalised muscle activity, dental occlusion, and mandibular movement analysis

This study aimed to predict dental freeway space by examining the clinical history, habits, occlusal parameters, mandibular hard tissue movement, soft tissue motion, muscle activity, and temporomandibular joint function of 66 participants. Data collection involved video-based facial landmark tracking, mandibular electrognathography, surface electromyography of mandibular range of motion, freeway space, chewing tasks, phonetic expressions, joint vibration analysis, and 3D jaw scans of occlusion. This resulted in a dataset of 121 predictor features, with freeway space as the target variable. Six models were trained on synthetic data ranging from 500 to 25,000 observations, with 65 original observations reserved for testing: Linear Regression, Random Forest, CatBoost Regressor, XGBoost Regressor, Multilayer Perceptron Neural Network (MPNN), and TabNet. Explainable AI indicated that key predictors of freeway space included phonetics, resting temporalis muscle activity, mandibular muscle activity during clenching, body weight, mandibular hard tissue lateral displacements, and dental arch parameters. CatBoost excelled with a test error of 0.65 mm using 5000 synthetic data points, while a refined MPNN achieved the best performance with 25,000 synthetic data points and 121 unique predictors, yielding an absolute error of 0.43 mm on the 65 original observations.

A New Perspective on the Homogeneous Coordinate System for Calculating Interatomic Distances and Their Derivatives in Terms of Internal Coordinates

AbstractIn many chemical physics calculations, efficient computation of interatomic distances and their derivatives using internal coordinates is essential for understanding molecular structure and dynamics. This paper explores the homogeneous coordinate system as an alternative framework to simplify these calculations. It is proven that computing interatomic distances using the Euclidean model of 3D space requires almost 60% more operations than the homogeneous model, which may represent a significant difference in molecular dynamics simulations.

Perturbation theory in the complete degenerate active space (CDAS-PT2).

Methods based on the multireference perturbation theory (MRPT) with the one-electron zeroth-order Hamiltonian are widely used for the description of excited states, for example, due to their relatively low computational cost. However, current methods have a common drawback-use of a model space with low size. In this article, we propose the MRPT method with the model space extended to the complete active space. The one-electron zeroth-order Hamiltonian suitable for this extension is formulated. The proposed method was applied to common models, such as LiF, ethylene, and trans-butadiene. It was shown to have accuracy superior to XMCQDPT2 in most cases, especially in the case of the small active space.

Study of the Nuclear Structure Properties in Strontium (90,92,94Sr) Isotopes Using Nuclear Shell-model Calculations

In the current research, various nuclear properties energy spectrum, reduced electromagnetic transition probabilities, nuclear moments, and the distributions of both the nuclear charge and mass density as a function of radial distance from the nucleus center (r) were computed for90,92,94Sr isotopes' using the NuShellX@MSU code. The Skyrme (SLy4) potential was utilized to compute the zirconium isotopes' wave functions with mass numbers 90, 92, and 94. By employing the Gloeckner interaction and bare G-matrix, the computed results showed good agreement with the available experimental information on the aforementioned nuclear features of all the above isotopes. Additionally, the spins and parities of energy levels were confirmed and determined in accordance with certain empirical values. Furthermore an acceptable agreement for transition strengthsB(E2;2_1^+→0_1^+ ) for 90,92,94Sr , and the dipole magnetic moment of the ground state in the90Sr isotope, was observed with the available experimental values. In these calculations, new values were predicted for the above nuclear properties, which had not been previously determined experimentally. KEYWORDS: energy spectra; gloeckner interaction; model space; nuShellX code; skyrme potential; transition strengths

Weight Decompositions on Algebraic Models for Mapping Spaces and Homotopy Automorphisms

We obtain restrictions on the rational homotopy types of mapping spaces and of classifying spaces of homotopy automorphisms by means of the theory of positive weight decompositions. The theory applies, in particular, to connected components of holomorphic maps between compact Kähler manifolds as well as homotopy automorphisms of Kähler manifolds.

Simulation-based Inference on Virtual Brain Models of Disorders

Abstract Connectome-based models, also known as Virtual Brain Models (VBMs), have been well established in network neuroscience to investigate pathophysiological causes underlying a large range of brain diseases. The integration of an individual's brain imaging data in VBMs has improved patient-specific predictivity, although Bayesian estimation of spatially distributed parameters remains challenging even with state-of-the-art Monte Carlo sampling. VBMs imply latent nonlinear state space models driven by noise and network input, necessitating advanced probabilistic machine learning techniques for widely applicable Bayesian estimation. Here we present Simulation-based Inference on Virtual Brain Models (SBI-VBMs), and demonstrate that training deep neural networks on both spatio-temporal and functional features allows for accurate estimation of generative parameters in brain disorders. The systematic use of brain stimulation provides an effective remedy for the non-identifiability issue in estimating the degradation limited to smaller subset of connections. By prioritizing model structure over data, we show that the hierarchical structure in SBI-VBMs renders the inference more effective, precise and biologically plausible. This approach could broadly advance precision medicine by enabling fast and reliable prediction of patient-specific brain disorders.

The African Penguin Spheniscus demersus should be considered Critically Endangered

The African Penguin Spheniscus demersus has been considered a threatened species since 1984 and, aside from a short period around 2000, its population has been in constant decline since at least the 1950s. By combining counts of the numbers of breeding pairs made at 26 colonies in South Africa and Namibia between 1979 and 2023 with Bayesian state–space models we reassess the species’ conservation status. The breeding population has declined by 77.8% (95% credible intervals: 71.8–84.6%) over the last 30 years (3 generations), from ∼44 300 breeding pairs in 1993 to ∼9 900 pairs in 2023. This falls just below the threshold for a global IUCN Red List status of Critically Endangered (CR) under criterion A2. However, the decline in Namibia exceeds that threshold at the national level (30-year decline = 82.4%, 78.2–86.2%) following the loss of ∼3 600 breeding pairs there in the last 5 years. In South Africa, the Western Cape population is now declining at <1.5% per annum, but the annual rate of decline in the Eastern Cape has worsened substantially, reaching 13% over the last 10 years. Overall, the global population has more than halved in the last decade and has fallen below 10 000 pairs for the first time. Moreover, “moving window reductions” over 3 generations using observed and projected population trajectories (up to the year 2033) indicate that the decline of the African Penguin population will exceed the 80% CR threshold (under criterion A4ab) with a high probability by 2028. Accordingly, we propose that the African Penguin should be transferred to the IUCN Red List category Critically Endangered.

Visual State Space Model for Image Deraining with Symmetrical Scanning

Image deraining aims to mitigate the adverse effects of rain streaks on image quality. Recently, the advent of convolutional neural networks (CNNs) and Vision Transformers (ViTs) has catalyzed substantial advancements in this field. However, these methods fail to effectively balance model efficiency and image deraining performance. In this paper, we propose an effective, locally enhanced visual state space model for image deraining, called DerainMamba. Specifically, we introduce a global-aware state space model to better capture long-range dependencies with linear complexity. In contrast to existing methods that utilize fixed unidirectional scan mechanisms, we propose a direction-aware symmetrical scanning module to enhance the feature capture of rain streak direction. Furthermore, we integrate a local-aware mixture of experts into our framework to mitigate local pixel forgetting, thereby enhancing the overall quality of high-resolution image reconstruction. Experimental results validate that the proposed method surpasses state-of-the-art approaches on six benchmark datasets.

Modelling systematic anatomical uncertainties of head and neck cancer patients during fractionated radiotherapy treatment.

Head and neck cancer patients experience systematic anatomical changes as well as random day to day anatomical changes during fractionated radiotherapy treatment. Modelling the expected systematic anatomical changes could aid in creating treatment plans which are more robust against such changes. A patient specific (SM) and population average (AM) model are presented which are able to capture the systematic anatomical changes of some head and neck cancer patients over the course of radiotherapy treatment. &#xD;Inter- patient correspondence aligned all patients to a model space. Intra- patient correspondence between each planning CT scan and on treatment cone beam CT scans was obtained using diffeomorphic deformable image registration. The stationary velocity fields were then used to develop B-Spline based SMs and AMs. The models were evaluated geometrically and dosimetrically. A leave-one-out method was used to compare the training and testing accuracy of the models. &#xD;Both SMs and AMs were able to capture systematic changes. The average surface distance between the registration propagated contours and the contours generated by the SM was less than 2mm, showing that the SM are able to capture the anatomical changes which a patient experiences during the course of radiotherapy. The testing accuracy was lower than the training accuracy of the SM, suggesting that the model overfits to the limited data available and therefore also captures some of the random day to day changes. For most patients the AMs were a better estimate of the anatomical changes than assuming there were no changes, but the AMs could not capture the variability in the anatomical changes seen in all patients. No difference was seen in the training and testing accuracy of the AMs. These observations were highlighted in both the geometric and dosimetric evaluations and comparisons. The large patient variability highlights the need for more complex, capable population models.

Modelling systematic anatomical uncertainties of head and neck cancer patients during fractionated radiotherapy treatment.

Head and neck cancer patients experience systematic anatomical changes as well as random day to day anatomical changes during fractionated radiotherapy treatment. Modelling the expected systematic anatomical changes could aid in creating treatment plans which are more robust against such changes. A patient specific (SM) and population average (AM) model are presented which are able to capture the systematic anatomical changes of some head and neck cancer patients over the course of radiotherapy treatment. &#xD;Inter- patient correspondence aligned all patients to a model space. Intra- patient correspondence between each planning CT scan and on treatment cone beam CT scans was obtained using diffeomorphic deformable image registration. The stationary velocity fields were then used to develop B-Spline based SMs and AMs. The models were evaluated geometrically and dosimetrically. A leave-one-out method was used to compare the training and testing accuracy of the models. &#xD;Both SMs and AMs were able to capture systematic changes. The average surface distance between the registration propagated contours and the contours generated by the SM was less than 2mm, showing that the SM are able to capture the anatomical changes which a patient experiences during the course of radiotherapy. The testing accuracy was lower than the training accuracy of the SM, suggesting that the model overfits to the limited data available and therefore also captures some of the random day to day changes. For most patients the AMs were a better estimate of the anatomical changes than assuming there were no changes, but the AMs could not capture the variability in the anatomical changes seen in all patients. No difference was seen in the training and testing accuracy of the AMs. These observations were highlighted in both the geometric and dosimetric evaluations and comparisons. The large patient variability highlights the need for more complex, capable population models.

Optimizing data acquisition: a Bayesian approach for efficient machine learning model training

Abstract Acquiring a substantial number of data points for training accurate machine learning (ML) models is a big challenge in scientific fields where data collection is resource-intensive. Here, we propose a novel approach for constructing a minimal yet highly informative database for training ML models in complex multi-dimensional parameter spaces. To achieve this, we mimic the underlying relation between the output and input parameters using Gaussian process regression (GPR). Using a set of known data, GPR provides predictive means and standard deviation for the unknown data. Given the predicted standard deviation by GPR, we select data points using Bayesian optimization to obtain an efficient database for training ML models. We compare the performance of ML models trained on databases obtained through this method, with databases obtained using traditional approaches. Our results demonstrate that the ML models trained on the database obtained using Bayesian optimization approach consistently outperform the other two databases, achieving high accuracy with a significantly smaller number of data points. Our work contributes to the resource-efficient collection of data in high-dimensional complex parameter spaces, to achieve high precision machine learning predictions.

Lepton flavour violation signal of the singly charged scalar singlet at the ILC

The singly charged SU(2) L singlet scalar is one of the very interesting new particles, as it can generate neutrino masses at loop level, produce contributions to various flavour observables. We study the possibility of detecting this kind of scalar predicted by the singly-charged scalar model at ILC via the lepton flavour violation process e+e−→S+S−→μe+E . Considering the constraints on the free parameters, we obtain the expected sensitivities of the ILC with the center of mass energy s=1TeV and the integrated luminosity L= 1.5 ab−1 to the parameter space of the singly-charged scalar model. The prospective excluded mass range at 95% C.L. is M S ≳ 470 GeV, 410 GeV for the branching ratio Bμe = 100% , 50%, respectively, while the scalar with M S ≳ 300 GeV is excluded at 95% C.L. for Bμe = 30%.

Topological consistency with low imperceptibility for graph adversarial attacks

Recent research shows that graph neural networks (GNNs) are easy to receive disruptions due to the lack of robustness, the phenomenon that poses a serious security threat. Currently, most efforts to attack GNNs mainly use gradient information to guide the attacks. However, the unreliability of gradient information, and the perceptibility of adversarial examples pose challenges that impede further progress in researching on graph adversarial attacks. From the unreliability of gradient information, we propose a Graph Distance Topological Consistency (GDTC). The scheme introduces graph connectivity, geodesic distance, cosine similarity, and Minkowski distance to construct the similarity matrices of the input space and embedding space of the surrogate model. The difference between the two similarities matrices is constrained during the training process of the surrogate model so that the surrogate model fully learns the topology of the original graph. From adversarial examples perceptibility, we propose attack loss with homogeneity restriction. Experiments show that GDTC can study the topological information of the original graph, enhance the reliability of gradient information, and significantly boost attack performance.

Topological consistency with low imperceptibility for graph adversarial attacks

Recent research shows that graph neural networks (GNNs) are easy to receive disruptions due to the lack of robustness, the phenomenon that poses a serious security threat. Currently, most efforts to attack GNNs mainly use gradient information to guide the attacks. However, the unreliability of gradient information, and the perceptibility of adversarial examples pose challenges that impede further progress in researching on graph adversarial attacks. From the unreliability of gradient information, we propose a Graph Distance Topological Consistency (GDTC). The scheme introduces graph connectivity, geodesic distance, cosine similarity, and Minkowski distance to construct the similarity matrices of the input space and embedding space of the surrogate model. The difference between the two similarities matrices is constrained during the training process of the surrogate model so that the surrogate model fully learns the topology of the original graph. From adversarial examples perceptibility, we propose attack loss with homogeneity restriction. Experiments show that GDTC can study the topological information of the original graph, enhance the reliability of gradient information, and significantly boost attack performance.

Structure-based out-of-distribution (OOD) materials property prediction: a benchmark study

In real-world materials research, machine learning (ML) models are usually expected to predict and discover novel exceptional materials that deviate from the known materials. It is thus a pressing question to provide an objective evaluation of ML model performances in property prediction of out-of-distribution (OOD) materials that are different from the training set. Traditional performance evaluation of materials property prediction models through the random splitting of the dataset frequently results in artificially high-performance assessments due to the inherent redundancy of typical material datasets. Here we present a comprehensive benchmark study of structure-based graph neural networks (GNNs) for extrapolative OOD materials property prediction. We formulate five different categories of OOD ML problems for three benchmark datasets from the MatBench study. Our extensive experiments show that current state-of-the-art GNN algorithms significantly underperform for the OOD property prediction tasks on average compared to their baselines in the MatBench study, demonstrating a crucial generalization gap in realistic material prediction tasks. We further examine the latent physical spaces of these GNN models and identify the sources of CGCNN, ALIGNN, and DeeperGATGNN’s significantly more robust OOD performance than those of the current best models in the MatBench study (coGN and coNGN) as a case study for the perovskites dataset, and provide insights to improve their performance.

PT -symmetry-enabled stable modes in a multicore fiber

Open systems with balanced gain and loss, described by parity-time (PT-symmetric) Hamiltonians have been deeply explored over the past decade. Most explorations are limited to finite discrete models (in real or reciprocal spaces) or continuum problems in one dimension. As a result, these models do not leverage the complexity and variability of two-dimensional continuum problems on a compact support. Here, we investigate eigenvalues of the Schrödinger equation on a disk with zero boundary condition, in the presence of constant, PT-symmetric, gain-loss potential that is confined to two mirror-symmetric disks. We find a rich variety of exceptional points, re-entrant PT-symmetric phases, and a nonmonotonic dependence of the PT-symmetry breaking threshold on the system parameters. By comparing results of two model variations, we show that this simple model of a multicore fiber supports propagating modes in the presence of gain and loss. Published by the American Physical Society 2024

Research on Zoning and Carbon Sink Enhancement Strategies for Ecological Spaces in Counties with Different Landform Types

Ecological carbon sinks, pivotal in mitigating carbon emissions, are indispensable for climate change mitigation. Counties, as the fundamental units of ecological space management, directly impact the achievement of regional dual carbon targets through their levels of carbon sink. However, existing research has overlooked the intricate relationship between terrain features and ecological spaces, leading to a lack of specific guidance on enhancing the carbon sink for counties with diverse landform characteristics. This study focused on Jingbian County (Loess Plateau), Fuping County (Guanzhong Plain), and Chenggu County (Qinba Mountains), each characterized by distinct landform characteristics. This study proposes a comprehensive identification model for ecological space within the context of dual carbon targets. Utilizing this model as a basis, the land use structure, carbon sink potential, and ecological spatial patterns of different counties were systematically analyzed. The results indicated substantial disparities in land use structure, carbon sink capabilities, and ecological space distributions among counties with different landform types. Specifically, Jingbian County was predominantly covered by grassland, exhibiting a moderate overall carbon sink capacity, with baseline ecological spaces playing a significant role. Conversely, Fuping County, dominated by cultivated land and construction land, exhibited the lowest carbon sink capacity, with non-ecological spaces accounting for a staggering 85.93%. Chenggu County, on the other hand, was characterized by the dominance of forestland, with nearly all its carbon sink originating from forestland, and core ecological spaces occupying a leading position. Tailored optimization strategies are recommended based on varying terrain features: Jingbian County should prioritize ecosystem restoration and conservation, while Fuping County should concentrate on optimizing land use structure and promoting urban greening. Reinforcing the carbon sink capacity of existing ecosystems is crucial for Chenggu County. This study broadens the perspective on ecological space optimization and provides scientific guidance and pragmatic insights tailored to regional disparities, which are instrumental in assisting various regions to achieve their dual carbon targets.

Linear control strategy of the dilution rate for stability in the anaerobic digestion process

Anaerobic digestion is a biochemical process of decomposition of organic matter, mediated by anaerobic bacteria, in an oxygen-free environment, from which biogas is produced. Biogas is used to generate thermal and electrical energy, where environmental conditions such as pH, temperature, concentration of volatile fatty acids, have a great influence on the efficiency and performance of the anaerobic digestion process, and can cause the process to be ruined. This document presents a control strategy that seeks to regulate the conditions of the anaerobic digestion process to keep it within the stability zone, using a linearization technique of the model in the state space for the design of controllers. In this work, a compensator is designed, for which the approximate linearization technique of the AM2 nonlinear model is proposed, to obtain the linearized model from which a classic lag-lead control strategy is proposed, using the dilution rate as a control variable. As a result, an improvement in system performance is obtained by reducing the maximum overshoot and reducing the settling time, evaluated at the operating point, which is validated through simulation and analysis of the response provided by the nonlinear model at the point of operation.