Inference Relations Research Articles

Active Spatial Positions Based Hierarchical Relation Inference for Group Activity Recognition

Group activity recognition aims to recognize behaviors characterized by multiple individuals within a scene. Existing schemes rely on individual relation inference and usually take the individuals as tokens. Essentially they select the most relevant region of the group activity from the entire image while filtering out irrelevant background noises. However, these schemes require individual bounding box labeling in both training and testing stages. Since individuals have usually been presented at one scale, multi-scale individuals cannot be combined in an effective way. In this paper, we present a novel end-to-end hierarchical relation inference framework based on active spatial positions for group activity recognition. This framework is designed to locate active spatial positions and use them as visual tokens to infer the relations for token embeddings. It requires individual bounding box labeling only in the training stage while automatically eliminating the background after locating active spatial positions from the entire scene. The hierarchical relations can be naturally inferred based on the visual tokens at different scales, contributing to further performance improvement. Experimental results demonstrate that the proposed framework is competitive against existing schemes that require more laboring and computation to generate labels in both the training and testing stage.

Seeing far: Abstract construal and visual distance judgments.

construal underlies mental travel. As a result, the human mind associates abstraction and psychological distance, whereby prompting abstract construal begets the inference of psychological distance - in time, social distance, hypotheticality, and physical space. That final distance is the only dimension that can be appraised visually, so would abstract construal impact judgments related to perceived visual distance? Two experiments provide evidence that abstract construal causes targets in the visual field to be judged as physically farther away. Further, the exacerbated sense of distance gives rise to related inferences about those visual targets (size and weight). These results deepen and broaden Construal Level Theory with practical implications for how people reason about the physical properties of objects - including but not limited to their physical distance.

Application of graph neural network and feature information enhancement in relation inference of sparse knowledge graph

At present, knowledge embedding methods are widely used in the field of knowledge graph (KG) reasoning, and have been successfully applied to those with large entities and relationships. However, in research and production environments, there are a large number of KGs with a small number of entities and relations, which are called sparse KGs. Limited by the performance of knowledge extraction methods or some other reasons (some common-sense information does not appear in the natural corpus), the relation between entities is often incomplete. To solve this problem, a method of the graph neural network and information enhancement is proposed. The improved method increases the mean reciprocal rank (MRR) and Hit@3 by 1.6% and 1.7%, respectively, when the sparsity of the FB15K-237 dataset is 10%. When the sparsity is 50%, the evaluation indexes MRR and Hit@10 are increased by 0.8% and 1.8%, respectively.

Concept-Enhanced Relation Network for Video Visual Relation Inference

Video visual relation inference aims at extracting the relation triplets in the form of < <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">subject-predicate-object</i> > in videos. With the development of deep learning, existing approaches are designed based on data-driven neural networks. But the datasets are always biased in terms of objects and relation triplets, which make relation inference challenging. Existing approaches often describe the relationships from visual, spatial, and semantic characteristics. The semantic description plays a key role to indicate the potential linguistic connections between objects, that are crucial to transfer knowledge across relationships, especially for the determination of novel relations. However, in these works, the semantic features are not emphasized, but simply obtained by mapping object labels, which can not reflect sufficient linguistic meanings. To alleviate the above issues, we propose a novel network, termed Concept-Enhanced Relation Network (CERN), to facilitate video visual relation inference. Thanks to the attributes and linguistic contexts implied in concepts, the semantic representations aggregated with related concept knowledge of objects are of benefit to relation inference. To this end, we incorporate retrieved concepts with local semantics of objects via the gating mechanism to generate the concept-enhanced semantic representations. Extensive experimental results show that our approach has achieved state-of-the-art performance on two public datasets: ImageNet-VidVRD and VidOR.

WITHDRAWN: Application of graph neural network and feature information enhancement in relation inference of sparse knowledge graph

WITHDRAWN: Application of graph neural network and feature information enhancement in relation inference of sparse knowledge graph

Bayesian inference of spatial and temporal relations in AI patents for EU countries

In the paper, we propose two models of Artificial Intelligence (AI) patents in European Union (EU) countries addressing spatial and temporal behaviour. In particular, the models can quantitatively describe the interaction between countries or explain the rapidly growing trends in AI patents. For spatial analysis Poisson regression is used to explain collaboration between a pair of countries measured by the number of common patents. Through Bayesian inference, we estimated the strengths of interactions between countries in the EU and the rest of the world. In particular, a significant lack of cooperation has been identified for some pairs of countries. Alternatively, an inhomogeneous Poisson process combined with the logistic curve growth accurately models the temporal behaviour by an accurate trend line. Bayesian analysis in the time domain revealed an upcoming slowdown in patenting intensity.

Impact of postoperative NSAIDs administration on anastomotic leak after esophago-gastric surgery: systematic review and meta-analysis.

Anastomotic leak (AL) is a feared complication of esophago-gastric surgery. Non-steroidal anti-inflammatory drugs (NSAIDs) are widely used to treat postoperative pain. Previous analyses conveyed heterogeneous data for colorectal surgery with a tendency toward high risk for AL after NSAIDs administration. In the setting of upper gastrointestinal (GI) surgery data are even more puzzled. The purpose of the present study was to assess whether an association exists between postoperative NSAIDs administration and AL after esophago-gastric surgery. PubMed, MEDLINE, Scopus, and Web of Science were searched up to November 2022. The included studies evaluated outcomes for NSAIDs vs. no NSAIDs administration after esophago-gastric surgery. The primary outcome was anastomotic leak (AL). Risk ratio (RR) and 95% confidence intervals (95% CI) were used to assess pooled effect size and relative inference. Six studies (43,784 patients) were included. The patient age ranged from 31 to 84years, 82.4% were males and preoperative BMI ranged from 15 to 31kg/m2. Esophagectomy was performed in 95% of patients. NSAIDs were administered in 18,075 (41.3%) patients. The cumulative incidence of AL was similar for NSAIDs vs. no NSAIDs (13.6% vs. 13.4%). The risk for postoperative AL was similar for NSAIDs vs. no NSAIDs administration (RR 1.49; 95% CI 0.81-2.75; p = 0.19). The cumulative incidence of postoperative gastrointestinal bleeding (0.36% vs. 0.39%), acute kidney injury (0.62% vs. 0.71%), and in-hospital mortality (2.39% vs. 2.66%) were comparable. NSAIDs administration for postoperative analgesia seems not associated with an increased risk for AL after esophago-gastric surgery.

Linear- Versus Circular-Stapled Esophagojejunostomy During Total Gastrectomy: Systematic Review and Meta-Analysis.

Background: While numerous techniques have been defined for esophagojejunostomy (EJ) during total gastrectomy including hand-sewn and stapled anastomoses, mechanical linear-stapled (LS) and circular-stapled (CS) anastomoses are widely adopted. However, there are scarce data on the optimal stapled technique for EJ during total gastrectomy. Materials and Methods: Scopus, Web of Science, MEDLINE, and PubMed were investigated up to October 30, 2022. We considered articles that appraised short-term outcomes after LS versus CS anastomosis in patients undergoing total gastrectomy for gastric cancer. Anastomotic leak (AL), anastomotic stricture (AS), and anastomotic bleeding (AB) were primary outcomes. Risk ratio (RR) and standardized mean difference (SMD) were used as pooled effect size measures, whereas 95% confidence intervals (95% CIs) were used to calculate related inference. Results: Sixteen studies (3156 patients) were incorporated. Overall, 1540 (48.8%) underwent CS, whereas 1616 (51.2%) underwent LS. Compared with CS, LS was related to a condensed RR for AS (RR: 0.27; 95% CI 0.15-0.49; P < .01), whereas no differences were found for AL (RR: 0.75; 95% CI 0.51-1.10; P = .14) and AB (RR: 0.59; 95% CI 0.24-1.44; P = .25). Postoperative pneumonia (RR: 0.98; P = .94), operative time (SMD: 0.51; P = .31), days to soft diet (SMD: -0.08; P = .36), hospital stay (SMD: 0.19; P = .46), and 30-day mortality (RR: 1.76; P = .31) were comparable between LS and CS. Conclusions: For EJ during total gastrectomy, our results suggest that LS seems related to a reduced risk of AS compared with CS, although no significant differences were found for the risk of AL and AB between the two techniques. Clinical Trial Registration number: CRD42022381221.

Unsupervised relational inference using masked reconstruction

Problem settingStochastic dynamical systems in which local interactions give rise to complex emerging phenomena are ubiquitous in nature and society. This work explores the problem of inferring the unknown interaction structure (represented as a graph) of such a system from measurements of its constituent agents or individual components (represented as nodes). We consider a setting where the underlying dynamical model is unknown and where different measurements (i.e., snapshots) may be independent (e.g., may stem from different experiments).MethodOur method is based on the observation that the temporal stochastic evolution manifests itself in local patterns. We show that we can exploit these patterns to infer the underlying graph by formulating a masked reconstruction task. Therefore, we propose GINA (Graph Inference Network Architecture), a machine learning approach to simultaneously learn the latent interaction graph and, conditioned on the interaction graph, the prediction of the (masked) state of a node based only on adjacent vertices. Our method is based on the hypothesis that the ground truth interaction graph—among all other potential graphs—allows us to predict the state of a node, given the states of its neighbors, with the highest accuracy.ResultsWe test this hypothesis and demonstrate GINA’s effectiveness on a wide range of interaction graphs and dynamical processes. We find that our paradigm allows to reconstruct the ground truth interaction graph in many cases and that GINA outperforms statistical and machine learning baseline on independent snapshots as well as on time series data.

Inferring biological kinship in ancient datasets: comparing the response of ancient DNA-specific software packages to low coverage data

BackgroundThe inference of biological relations between individuals is fundamental to understanding past human societies. Caregiving, resource sharing and sexual behaviours are often mediated by biological kinship and yet the identification and interpretation of kin relationships in prehistoric human groups is difficult. In recent years, the advent of archaeogenetic techniques have offered a fresh approach, and when combined with more traditional osteological and interpretive archaeological methods, allows for improved interpretation of the burial practices, cultural behaviours, and societal stratification in ancient societies. Although archaeogenetic techniques are developing at pace, questions remain as to their accuracy, particularly when applied to the low coverage datasets that results from the sequencing of DNA derived from highly degraded ancient material.ResultsThe performance of six of the most commonly used kinship identifcation software methods was explored at a range of low and ultra low genome coverages. An asymmetrical response was observed across packages, with decreased genome coverage resulting in differences in both direction and degree of change of calculated kinship scores and thus pairwise relatedness estimates are dependant on both package used and genome coverage. Methods reliant upon genotype likelihoods methods (lcMLkin, NGSrelate and NGSremix) show a decreased level of prediction at coverage below 1x, although were consistent in the particular relationships identified at these coverages when compared to the pseudohaploid reliant methods tested (READ, the Kennett 2017 method and TKGWV2.0). The three pseudohaploid methods show predictive potential at coverages as low as 0.05x, although the accuracy of the relationships identified is questionable given the increase in the number of relationships identifIed at the low coverage (type I errors).ConclusionTwo pseudohaploid methods (READ and Kennett 2017) show relatively consistent inference of kin relationships at low coverage (0.5x), with READ only showing a significant performance drop off at ultralow coverages (< 0.2x). More generally, our results reveal asymmetrical kinship classifications in some software packages even at high coverages, highlighting the importance of applying multiple methods to authenticate kin relationships in ancient material, along with the continuing need to develop laboratory methods that maximise data output for downstream analyses.

Bayesian Analysis of Spatial Model for Frequency of Tornadoes

Frequency analysis of tornadoes is very important for risk analysis and disaster control. In this paper, the annual frequency of tornadoes that occurred in the United States from 1967 to 2016 is analyzed. The simple analysis shows that frequencies of tornadoes of different sites are spatially correlated and over-dispersed. To explain the two characteristics of the data, the Bayesian hierarchical model is proposed. For comparison purposes, the Bayesian model with negative binomial distribution, Poisson distribution, Polya distribution, and first-order, non-negative, integer-valued autoregressive model with Bell innovations(BL-INAR(1)) are considered to fit the frequency of tornado. The distribution parameters of all sites are assumed to be spatially correlated, and the corresponding Bayesian hierarchical models were established. MCMC (Markov Chain Monte Carlo) method is applied to parameter estimations and relative statistical inference. By comparison of the analysis results, the negative binomial distribution is recommended to analyze the overdispersion and spatial correlation among the sites of the data. The comparison between the simulated frequencies based on the proposed model and the actual frequencies also verifies that the proposed method is a better model for the data.

Inference of Factors for Labor Productivity Growth Used Randomized Experiment and Statistical Causality

The study of causal dependencies in economics is fraught with great difficulties, that it is required to consider not only the object structure, but also take into account a huge number of factors acting on the object, about which nothing is either known or difficult to measure. In this paper, we attempt to overcome this problem and apply the theory of statistical causality for labor productivity management. We suggest new technology that provides the inference of causal relations between the special programs implemented in the company’s and employee’s labor productivity. The novelty of the proposed technology is that it is based on a hybrid object model, combines two models: 1—the structural object model about its functioning and development to provide a causal inference and prediction the effect of explicit factors; 2—the model based on observed data to clarify causality and to test it empirically. The technology provides integration of the theory of causal Bayesian networks, methods of randomized controlled experiments and statistical methods, allows under nonlinearity, dynamism, stochasticity and non-stationarity of the initial data, to evaluate the effect of programs on the labor effeciency. The difference between the proposed technology and others is that it ensures determination the synergistic effect of the action of the cause (program) on the effect—labor productivity in condition of hidden factors. The practical significance of the research is the results of its testing the proposed theoretical provisions, methods and technologies on actual data about food service company. The results obtained could contribute to the labor productivity growth over uncertainty of the external and internal factors and provide the companies sustainable development and its profitability growth.

Graph Representation Learning Meets Computer Vision: A Survey

A graph structure is a powerful mathematical abstraction, which can not only represent information about individuals but also capture the interactions between individuals for reasoning. Geometric modeling and relational inference based on graph data is a long-standing topic of interest in the computer vision community. In this article, we provide a systematic review of graph representation learning and its applications in computer vision. First, we sort out the evolution of representation learning on graphs, categorizing them into the nonneural network and neural network methods based on the way the nodes are encoded. Specifically, nonneural network methods, such as graph embedding and probabilistic graphical models, are introduced, and neural network methods, such as graph recurrent neural networks, graph convolutional networks, and variants of graph neural networks, are also presented. Then, we organize the applications of graph representation algorithms in various vision tasks (such as image classification, semantic segmentation, object detection, and tracking) for review and reference, and the typical graph construction approaches in computer vision are also summarized. Finally, on the background of biology and brain inspiration, we discuss the existing challenges and future directions of graph representation learning and computer vision.

Efficacy of laparoscopic Toupet fundoplication compared to endoscopic and surgical procedures for GERD treatment: a randomized trials network meta-analysis.

We aim to review and quantitatively compare laparoscopic Toupet fundoplication (LTF), Nissen fundoplication (LNF), anterior partial fundoplication (APF), magnetic augmentation sphincter (MSA), radiofrequency ablation (RFA), transoral incisionless fundoplication (TIF), proton pump inhibitor (PPI), and placebo for the treatment of GERD. A number of meta-analyses compared the efficacy of surgical and endoscopic procedures for recalcitrant GERD, but considerable debate on the effectiveness of operative strategies remains. A systematic review of MEDLINE databases, EMBASE, and Web of Science for randomized controlled trials (RCTs) comparing the aforementioned surgical and endoscopic GERD treatments was performed. Risk ratio and weighted mean difference were used as pooled effect size measures, whereas 95% credible intervals (CrI) were used to assess relative inference. Thirty-three RCTs were included. Surgical and endoscopic treatments have similar RR for heartburn, regurgitation, bloating. LTF has a lower RR of post-operative dysphagia when compared to APF (RR 3.3; Crl 1.4-7.1) and LNF (RR 2.5; Crl 1.3-4.4). The pooled network meta-analysis did not observe any significant improvement regarding LES pressure and pH < from baseline. LTF, APF, LNF, MSA, RFA, and TIF had have a similar post-operative PPI discontinuation rate. LTF has a lower rate of post-operative dysphagia when compared to APF and LNF. The pre-post effects, such as GERD-HQRL, LES pressure, and pH <4, should be avoided in meta-analyses because results may be biased. Last, a consensus about the evaluation of GERD treatments' efficacy and their outcomes is needed.

SENSIPLUS-LM: A Low-Cost EIS-Enabled Microchip Enhanced with an Open-Source Tiny Machine Learning Toolchain

The technological step towards sensors’ miniaturization, low-cost platforms, and evolved communication paradigms is rapidly moving the monitoring and computation tasks to the edge, causing the joint use of the Internet of Things (IoT) and machine learning (ML) to be massively employed. Edge devices are often composed of sensors and actuators, and their behavior depends on the relative rapid inference of specific conditions. Therefore, the computation and decision-making processes become obsolete and ineffective by communicating raw data and leaving them to a centralized system. This paper responds to this need by proposing an integrated architecture, able to host both the sensing part and the learning and classifying mechanisms, empowered by ML, directly on board and thus able to overcome some of the limitations presented by off-the-shelf solutions. The presented system is based on a proprietary platform named SENSIPLUS, a multi-sensor device especially devoted to performing electrical impedance spectroscopy (EIS) on a wide frequency interval. The measurement acquisition, data processing, and embedded classification techniques are supported by a system capable of generating and compiling code automatically, which uses a toolchain to run inference routines on the edge. As a case study, the system capabilities of such a platform in this work are exploited for water quality assessment. The joint system, composed of the measurement platform and the developed toolchain, is named SENSIPLUS-LM, standing for SENSIPLUS learning machine. The introduction of the toolchain empowers the SENSIPLUS platform moving the inference phase of the machine learning algorithm to the edge, thus limiting the needs of external computing platforms. The software part, i.e., the developed toolchain, is available for free download from GitLab, as reported in this paper.

Relative contrast estimation and inference for treatment recommendation.

When there are resource constraints, it may be necessary to rank individualized treatment benefits to facilitate the prioritization of assigning different treatments. Most existing literature on individualized treatment rules targets absolute conditional treatment effect differences as a metric for the benefit. However, there can be settings where relative differences may better represent such benefit. In this paper, we consider modeling such relative differences formed as scale-invariant contrasts between the conditional treatment effects. By showing that all scale-invariant contrasts are monotonic transformations of each other, we posit a single index model for a particular relative contrast. We then characterize semiparametric estimating equations, including the efficient score, to estimate index parameters. To achieve semiparametric efficiency, we propose a two-step approach that minimizes a doubly robust loss function for initial estimation and then performs a one-step efficiency augmentation procedure. Careful theoretical and numerical studies are provided to show the superiority of our proposedapproach.

Acoustic surveys improve landscape-scale detection of a critically endangered Australian bird, the plains-wanderer (Pedionomus torquatus)

Context Monitoring the population dynamics of threatened species requires a landscape-scale understanding of their distribution over time. However, detectability is inherently low for rare, widely dispersed, and cryptic species. For animals that vocalise, passive acoustic recorders allow for efficient and repeated surveys over a large geographic area, increasing inference in relation to detectability and occupancy. Aims Our aim was to determine how well acoustic surveys, combined with automated species detection, identified the presence of the critically endangered plains-wanderer (Pedionomus torquatus) relative to a traditional method of nocturnal spotlighting surveys at sites across the Northern Plains of Victoria, Australia. Methods Using Hidden Markov Models, we created 17 different plains-wanderer call recognisers by varying input parameters and assessed their performance on the same training and testing audio dataset. We then applied our best-performing recogniser to a field audio dataset to estimate detectability and compared the presence of plains-wanderers at sites paired with nocturnal surveys. Key results Recognisers varied in their overall performance in detecting individual plains-wanderer calls but were equally effective at determining whether any plains-wanderer calls were detected at a site within our training and testing datasets. Although survey effort was not standardised across field survey methods, we found audio surveys and nocturnal spotlight surveys were equally successful at establishing site-level occupancy; however, acoustic surveys provide the potential to survey more sites over a given time period. Conclusions We suggest acoustic surveys can be an effective and efficient means to document occupancy at the landscape scale, facilitating prioritisation of nocturnal surveys to assess population demographic parameters including abundance and breeding status. Implications Acoustic surveys can provide a complementary method to establish occupancy for cryptic, vocally active, threatened species. We provide recommendations on ways to develop an effective acoustic monitoring program workflow, from data collection to acoustic analysis, that can be used by different user groups.

Study on the allosteric activation mechanism of SHP2 via elastic network models and neural relational inference molecular dynamics simulation.

As a ubiquitous protein tyrosine phosphatase, SHP2 is involved in PD-1/PD-L1 mediated tumor immune escape and undergoes substantial conformational changes. Therefore, it is considered an ideal target for tumor intervention. However, the allosteric mechanisms of SHP2 binding PD-1 intracellular ITIM/ITSM phosphopeptides remain unclear, which greatly hinders the development of novel structure-based anticancer allosteric inhibitors. In this work, the open and closed structural models of SHP2 are first constructed based on this knowledge; next their motion modes are investigated via elastic network models such as the Gaussian network model (GNM), anisotropic network model (ANM) and adaptive anisotropic network model (aANM); and finally, a possible allosteric signaling pathway is proposed using a neural relational inference molecular dynamics (NRI-MD) simulation embedded with an artificial intelligence (AI) strategy. In GNM and ANM, the N-SH2, C-SH2 and PTP domains all exhibit distinct dynamics partitions, and the N-SH2/C-SH2 regions show a rigid rotation relative to PTP. According to a series of intermediate snapshots given by aANM, N-SH2 is first identified with pY223 specifically, inducing a D'E-loop to change from β-sheets to random coils, and then, C-SH2 serves as a fulcrum to drive N-SH2 to rotate 110° completely away from the original active sites of PTP. Finally, a possible allosteric signaling-transfer path for SHP2, namely R220-R138-T108-R32, is proposed based on NRI-MD sampling. This work provides a possible allosteric mechanism of SHP2, which is helpful for the following design of novel allosteric inhibitors and is expected to be used in clinical synergies with PD-1 monoclonal antibody.

AlignVE: Visual Entailment Recognition Based on Alignment Relations

Visual entailment (VE) is to recognize whether the semantics of a hypothesis text can be inferred from the given premise image, which is one special task among recent emerged vision and language understanding tasks. Currently, most of the existing VE approaches are derived from the methods of visual question answering. They recognize visual entailment by quantifying the similarity between the hypothesis and premise in the content semantic features from multi modalities. Such approaches, however, ignore the VE's unique nature of relation inference between the premise and hypothesis. Therefore, in this paper, a new architecture called AlignVE is proposed to solve the visual entailment problem with a relation interaction method. It models the relation between the premise and hypothesis as an alignment matrix. Then it introduces a pooling operation to get feature vectors with a fixed size. Finally, it goes through the fully-connected layer and normalization layer to complete the classification. Experiments show that our alignment-based architecture reaches 72.45\% accuracy on SNLI-VE dataset, outperforming previous content-based models under the same settings.

Exploring Sparse Spatial Relation in Graph Inference for Text-Based VQA.

Text-based visual question answering (TextVQA) faces the significant challenge of avoiding redundant relational inference. To be specific, a large number of detected objects and optical character recognition (OCR) tokens result in rich visual relationships. Existing works take all visual relationships into account for answer prediction. However, there are three observations: (1) a single subject in the images can be easily detected as multiple objects with distinct bounding boxes (considered repetitive objects). The associations between these repetitive objects are superfluous for answer reasoning; (2) two spatially distant OCR tokens detected in the image frequently have weak semantic dependencies for answer reasoning; and (3) the co-existence of nearby objects and tokens may be indicative of important visual cues for predicting answers. Rather than utilizing all of them for answer prediction, we make an effort to identify the most important connections or eliminate redundant ones. We propose a sparse spatial graph network (SSGN) that introduces a spatially aware relation pruning technique to this task. As spatial factors for relation measurement, we employ spatial distance, geometric dimension, overlap area, and DIoU for spatially aware pruning. We consider three visual relationships for graph learning: object-object, OCR-OCR tokens, and object-OCR token relationships. SSGN is a progressive graph learning architecture that verifies the pivotal relations in the correlated object-token sparse graph, and then in the respective object-based sparse graph and token-based sparse graph. Experiment results on TextVQA and ST-VQA datasets demonstrate that SSGN achieves promising performances. And some visualization results further demonstrate the interpretability of our method.