Continuous Space Representations Research Articles

RoCS: Knowledge Graph Embedding Based on Joint Cosine Similarity

Knowledge graphs usually have many missing links, and predicting the relationships between entities has become a hot research topic in recent years. Knowledge graph embedding research maps entities and relations to a low-dimensional continuous space representation to predict links between entities. The present research shows that the key to the knowledge graph embedding approach is the design of scoring functions. According to the scoring function, knowledge graph embedding methods can be classified into dot product models and distance models. We find that the triple scores obtained using the dot product model or the distance model were unbounded, which leads to large variance. In this paper, we propose RotatE Cosine Similarity (RoCS), a method to compute the joint cosine similarity of complex vectors as a scoring function to make the triple scores bounded. Our approach combines the rotational properties of the complex vector embedding model RotatE to model complex relational patterns. The experimental results demonstrate that the newly introduced RoCS yields substantial enhancements compared to RotatE across various knowledge graph benchmarks, improving up to 4.0% in hits at 1 (Hits@1) on WN18RR and improving up to 3.3% in Hits@1 on FB15K-237. Meanwhile, our method achieves some new state-of-the-art (SOTA), including Hits@3 of 95.6%, Hits@10 of 96.4% on WN18, and mean reciprocal rank (MRR) of 48.9% and Hits@1 of 44.5% on WN18RR.

Two-Step Asynchronous Iterative Formation Control for Heterogeneous Vehicles in Highway Scenarios

In this article a novel algorithm for the on-line control of formations of heterogeneous vehicles is proposed for highway traffic scenarios. A two-step iterative strategy relying on both a discrete and a continuous space representation, using Dynamic Programming and reference tracking via Model Predictive Control, is formulated. It allows the creation and arbitrary reshaping of vehicle formations in a unified context, providing an applicability in a wide range of practical situations. Only position measurements and basic communication capabilities for global coordination are required for the automated vehicles. Scalability and flexibility are achieved by the underlying decoupled structure of the algorithm, while an asynchronous functioning is enforced due to trigger signals that are exchanged between the vehicles and the coordinator. Different scenarios are simulated and discussed to evidence the effectiveness of the proposed strategy.

Biobjective route planning of an unmanned air vehicle in continuous space

We consider the route planning problem of an unmanned air vehicle (UAV) in a continuous space that is monitored by radars. The UAV visits multiple targets and returns to the base. The routes are constructed considering the total distance traveled and the total radar detection threat objectives. The UAV is capable of moving to any point in the terrain. This leads to infinitely many efficient trajectories between target pairs and infinitely many efficient routes to visit all targets. We use a two stage approach in solving the complex problem of finding all efficient routes. In the first stage, we structure the nondominated frontiers of the efficient trajectories between all target pairs. For this, we first identify properties shared by efficient trajectories between target pairs that are protected by a radar. This helps to structure the nondominated frontier between any target pair by identifying at most four specific efficient trajectories. We develop a search-based algorithm that finds these efficient trajectories effectively. For the second stage, we develop a mixed integer nonlinear program that exploits the structured nondominated frontiers between target pairs to construct the efficient routes. We compare the nondominated front we generate in the continuous space with its counterpart in a terrain discretized with three different grid fidelities. The continuous space representation outperforms all discrete representations in terms of solution quality and computational times.

A survey on discrete space and continuous space facility layout problems

PurposeThe purpose of this paper is to review, evaluate and classify the academic research that has been published in facility layout problems (FLPs) and to analyse how researches and practices on FLPs are.Design/methodology/approachThe review is based on 166 papers published from 1953 to 2021 in international peer-reviewed journals. The literature review on FLPs is presented under broader headings of discrete space and continuous space FLPs. The important formulations of FLPs under static and dynamic environments represented in the discrete and continuous space are presented. The articles reported in the literature on various representations of facilities for the continuous space Unequal Area Facility Layout Problems (UA-FLPs) are summarized. Discussed and commented on adaptive and robust approaches for dynamic environment FLPs. Highlighted the application of meta-heuristic solution methods for FLPs of a larger size.FindingsIt is found that most of the earlier research adopted the discrete space for the formulation of FLPs. This type of space representation for FLPs mostly assumes an equal area for all facilities. UA-FLPs represented in discrete space yield irregular shape facilities. It is also observed that the recent works consider the UA-FLPs in continuous space. The solution of continuous space UA-FLPs is more accurate and realistic. Some of the recent works on UA-FLPs consider the flexible bay structure (FBS) due to its advantages over the other representations. FBS helps the proper design of aisle structure in the detailed layout plan. Further, the recent articles reported in the literature consider the dynamic environment for both equal and unequal area FLPs to cope with the changing market environment. It is also found that FLPs are Non-deterministic Polynomial-complete problems, and hence, they set the challenges to researchers to develop efficient meta-heuristic methods to solve the bigger size FLPs in a reasonable time.Research limitations/implicationsDue to the extremely large number of papers on FLPs, a few papers may have inadvertently been missed. The facility layout design research domain is extremely vast which covers other areas such as cellular layouts, pick and drop points and aisle structure design. This research review on FLPs did not consider the papers published on cellular layouts, pick and drop points and aisle structure design. Despite the possibility of not being all-inclusive, the authors firmly believe that most of the papers published on FLPs are covered and the general picture presented on various approaches and parameters of FLPs in this paper are precise and trustworthy.Originality/valueTo the best of the authors’ knowledge, this paper reviews and classifies the literature on FLPs for the first time under the broader headings of discrete space and continuous space representations. Many important formulations of FLPs under static and dynamic environments represented in the discrete and continuous space are presented. This paper also provides the observations from the literature review and identifies the prospective future directions.

Data-driven framework for the adaptive exit selection problem in pedestrian flow: Visual information based heuristics approach

Pedestrian behavior during evacuation has been formulated using various arbitrary microscopic methods to investigate the performance of crowd dynamics while their custom rules result in low visual realism in simulation due to the complexity of intrinsic decision logic of human. Statistical analysis is an effective way to reveal the motion pattern and path planning behavior of pedestrians whose main idea is to approach the trajectory and social attributes data of pedestrians extracted from evacuation drills as much as possible. In this study, we present a data-driven based microscopic pedestrian-simulation model with continuous-space representation to explore the potential of integrating empirical analysis into crowd simulation to enhance the authenticity of decision making. This method extracts the pedestrian’s decision mode and smoothly applies it in the crowd dynamics model. Instead of navigating agents by arbitrary regulations, the desired direction of pedestrians during the motion is arranged by machine learning (ML) algorithms. The path decision module trained with actual pedestrian data improves the compatibility of the model in the application of various spatial scenarios and no longer suffers from tedious parameter fine-tuning work. To completely describe the information precepted by pedestrians, a polygon segmentation module is developed to divide the visual field of pedestrians and identify the mutual visibility among them. This module filters out the information that can be perceived by pedestrians in real situations, thereby bridges the gap between statistical analysis and numerical simulation methods. We compare different ML approaches for route-choice behavior prediction and discuss the relative importance of its influencing variables under different scenarios. Inferring the perception of social interactions from disaggregate choice data, the scope of effectiveness of conformity behavior and crowd-aversion are also discussed. The simulation results are compared with experimental data, illustrating the model’s capability to accurately reproduce the observed flow motion in various scenarios with moderate modification in physical environment initialization.

Route choice in the pedestrian evacuation: Microscopic formulation based on visual information

A pedestrian’s visual field constantly changes as they move around a room containing structural components or large furniture, which can block the pedestrian from perceiving visual information. Changes in position simultaneously update optional route sets and the perception of how crowded these paths are. In this study, a microscopic pedestrian–simulation model with continuous-space representation is developed to investigate the route-choice behavior of pedestrians in a space where obstacles block the visual field. Instead of potential-based navigation, the desired direction of pedestrians during the motion is expressed by an oriented network and delivered to the navigation layer: an optimal reciprocal collision avoidance (ORCA) system, which models pedestrians’ local maneuvers and movement decisions in detail. We quantify the characteristics of the model by comparing the model results with two groups of experiments conducted under conditions of good, limited, and zero visibility. The performance results with two different layouts are compared through numerical simulation, illustrating the model’s capability to accurately represent the observed crowd dynamics in the facility. Results from the sensitivity analysis and simulation scenario can benefit the evacuation guidance arrangement and internal component design in large public spaces.

Network-Based Continuous Space Representation for Describing Pedestrian Movement in High Resolution

A concept of network-based continuous space representation is proposed and applied to the sequential map matching problem with simulation data assuming pedestrian movement. The concept allows for dealing with situations that the resolution of network representation is not high enough to describe the pedestrian movement considering the observation accuracy. The experiment showed that the proposed concept worked well in the example of pedestrian movement along with the sidewalk by estimation of accurate positions.

Continuous space models for CLIR

We present and evaluate a novel technique for learning cross-lingual continuous space models to aid cross-language information retrieval (CLIR). Our model, which is referred to as external-data composition neural network (XCNN), is based on a composition function that is implemented on top of a deep neural network that provides a distributed learning framework. Different from most existing models, which rely only on available parallel data for training, our learning framework provides a natural way to exploit monolingual data and its associated relevance metadata for learning continuous space representations of language. Cross-language extensions of the obtained models can then be trained by using a small set of parallel data. This property is very helpful for resource-poor languages, therefore, we carry out experiments on the English-Hindi language pair. On the conducted comparative evaluation, the proposed model is shown to outperform state-of-the-art continuous space models with statistically significant margin on two different tasks: parallel sentence retrieval and ad-hoc retrieval.

Cross-language plagiarism detection over continuous-space- and knowledge graph-based representations of language

Cross-language (CL) plagiarism detection aims at detecting plagiarised fragments of text among documents in different languages. The main research question of this work is on whether knowledge graph representations and continuous space representations can complement to each other and improve the state-of-the-art performance in CL plagiarism detection methods. In this sense, we propose and evaluate hybrid models to assess the semantic similarity of two segments of text in different languages. The proposed hybrid models combine knowledge graph representations with continuous space representations aiming at exploiting their complementarity in capturing different aspects of cross-lingual similarity. We also present the continuous word alignment-based similarity analysis, a new model to estimate similarity between text fragments. We compare the aforementioned approaches with several state-of-the-art models in the task of CL plagiarism detection and study their performance in detecting different length and obfuscation types of plagiarism cases. We conduct experiments over Spanish-English and German-English datasets. Experimental results show that continuous representations allow the continuous word alignment-based similarity analysis model to obtain competitive results and the knowledge-based document similarity model to outperform the state-of-the-art in CL plagiarism detection.

A Sparse Plus Low-Rank Exponential Language Model for Limited Resource Scenarios

This paper describes a new exponential language model that decomposes the model parameters into one or more low-rank matrices that learn regularities in the training data and one or more sparse matrices that learn exceptions (e.g., keywords). The low-rank matrices induce continuous-space representations of words and histories. The sparse matrices learn multiword lexical items and topic/domain idiosyncrasies. This model generalizes the standard l1-regularized exponential language model, and has an efficient accelerated first-order training algorithm. Language modeling experiments show that the approach is useful in scenarios with limited training data, including low resource languages and domain adaptation.

From time representation in scheduling to the solution of strip packing problems

We propose two mixed-integer linear programming based approaches for the 2D orthogonal strip packing problem. Using knowledge from the alternative forms of time representation in scheduling formulations, we show how to efficiently combine three different concepts into the x- and y-dimensions. One model features a discrete representation on the x-axis (strip width) and a continuous representation with general precedence variables on the y-axis (strip height). The other features a full continuous-space representation with the same approach for the y-axis and a single non-uniform grid made up of slots for the x-axis. Through the solution of a set of twenty nine instances from the literature, we show that the former is a better approach, even when compared to three alternative MILP models ranging from a pure discrete-space to a pure continuous-space with precedence variables in both dimensions. All models are available in www.minlp.org.

Formulation of pedestrian movement in microscopic models with continuous space representation

When the microscopic pedestrian models, in which pedestrian space is continuously represented, are used to simulate pedestrian movement in the buildings with internal obstacles, some issues arise and need be dealt with in detail. This paper discusses two of the issues, namely formulating the desired direction of each pedestrian in the buildings and determining the region around each pedestrian, other individuals and obstacles in which affect his or her movement. The methods for computing the desired direction and effect region are proposed, using the algorithms for the potential of pedestrian space. By numerical experiments, the performance results of three proposed formulae for the desired direction are compared, the method for the effect region is tested, and the validity of the method for computing the desired direction as considering the border effect of obstacles is verified. Numerical results indicate that the proposed methods can be used to formulate pedestrian movement, especially in the buildings with internal obstacles, in the microscopic models with continuous space representation.

Simplified Computation for Nonparametric Windows Method of Probability Density Function Estimation

Recently, Kadir and Brady proposed a method for estimating probability density functions (PDFs) for digital signals which they call the Nonparametric (NP) Windows method. The method involves constructing a continuous space representation of the discrete space and sampled signal by using a suitable interpolation method. NP Windows requires only a small number of observed signal samples to estimate the PDF and is completely data driven. In this short paper, we first develop analytical formulae to obtain the NP Windows PDF estimates for 1D, 2D, and 3D signals, for different interpolation methods. We then show that the original procedure to calculate the PDF estimate can be significantly simplified and made computationally more efficient by a judicious choice of the frame of reference. We have also outlined specific algorithmic details of the procedures enabling quick implementation. Our reformulation of the original concept has directly demonstrated a close link between the NP Windows method and the Kernel Density Estimator.

Graphical Representations of Protein Sequences for Alignment-Free Comparative and Predictive Studies. Recognition of Protease Inhibition Pattern from H-Depleted Molecular Graph Representation of Protease Sequences

Biomacromolecular information is hinged by sequence and structure representations. Because structure is often more conserved than sequence, achieving function inference from structural similarity is easier than from sequence analysis. However, structural information is sparse and only available for a small part of the protein space. Detecting subtle similarities between proteins from sequence depends strongly on the representations used. Continuous-space representations yield promising results in comparative evolution analysis, structural classification and sequence-function/property relationship studies. These simple methods provide a pre-classification and/or feature generation stages to sophisticated classification methods. We review the state-of-the-art in protein sequence graphical representations along with some derived metrics for statistical pattern recognition analysis. In addition, the binding stability pattern of protease-inhibitor complexes is modelled from H-depleted molecular graph representation of protease sequences and ligands using support vector machines with about 80% prediction accuracy. Keywords: Chaos game representation, protein graph, QSAR analysis, pseudo-folding representation.

Developing a Continuous Space Representation of a Simulated Population

Spatial microsimulation models typically match census of population data with survey data in order to simulate synthetic populations of individuals and households within small-scale geographic areas. For most spatial microsimulation applications this level of spatial precision is satisfactory. For others, more precise information on the location of simulated units may be required. To this end this paper develops a continuous space representation of a simulated population. It presents a statistical matching approach for assigning simulated households from a spatial microsimulation model to unique spatially-referenced residential locations. The allocation is based on a random assignment after splitting the simulated households into two groups: those predicted to reside in apartments and those predicted to reside in houses. The resulting ‘geohouseholds’ have a range of potential applications in economic and spatial analysis. Création d'une représentation spatiale continue d'une population stimulée Résumé Les modèles de microsimulation spatiale assortissent généralement les données de recensement de la population à des données de sondages, afin de simuler des populations synthétiques de particuliers et de foyers au sein de régions géographiques à échelle restreinte. Dans la plupart des applications de microsimulation spatiale, ce niveau de précision spatiale est satisfaisant. Dans d'autres, des informations plus précises sur l'emplacement d'unités simulées pourront s'avérer nécessaires. A cette fin, la présente communication crée une représentation spatiale continue d'une population simulée. Elle présente une méthode de correspondance statistique permettant d'affecter des foyers simulés, issus d'un modèle de microsimulation spatiale à des lieux résidentiels unique à référence spatiale. Cette allocation est basée sur une affectation aléatoire après la subdivision des foyers simulés en deux groupes : ceux dont on prévoit qu'ils résideront en appartement, et ceux dont on prévoit qu'ils résideront dans un maison. Les « géofoyers » résultants présentent toute une série d'applications potentielles pour les analyses économiques et spatiales. Desarrollo de una representación espacial continua de una población simulada Extracto Típicamente, los modelos de microsimulación espacial emparejan el censo de datos de la población con datos de encuestas, con objeto de simular poblaciones sintéticas de individuos y hogares dentro de áreas geográficas a pequeña escala. Para la mayoría de las aplicaciones de microsimulación espacial este nivel de precisión espacial es satisfactorio. Para otras, podría requerirse información más precisa sobre la ubicación de unidades simuladas. Con este objetivo, este trabajo desarrolla la representación espacial continua de una población simulada. Presenta un planteamiento de emparejamiento estadístico para asignar hogares simulados procedentes de un modelo de microsimulación espacial a ubicaciones residenciales únicas referenciadas espacialmente. La colocación se basa en una asignación al azar después de dividir los hogares simulados en dos grupos: los que se predice que residirán en apartamentos y los que se predice que residirán en casas. Los ‘geohogares’ resultantes ofrecen una gama de aplicaciones en potencia en el análisis económico y espacial.

Constraint-based sensor planning for scene modeling

We describe an automated scene modeling system that consists of two components operating in an interleaved fashion: an incremental modeler that builds solid models from range imagery; and a sensor planner that analyzes the resulting model and computes the next sensor position. This planning component is target-driven and computes sensor positions using model information about the imaged surfaces and the unexplored space in a scene. The method is shape-independent and uses a continuous-space representation that preserves the accuracy of sensed data. It is able to completely acquire a scene by repeatedly planning sensor positions, utilizing a partial model to determine volumes of visibility for contiguous areas of unexplored scene. These visibility volumes are combined with sensor placement constraints to compute sets of occlusion-free sensor positions that are guaranteed to improve the quality of the model. We show results for the acquisition of a scene that includes multiple, distinct objects with high occlusion.