Diverse Granularity Research Articles

A unified framework for unsupervised action learning via global-to-local motion transformer

Human action recognition remains challenging due to the inherent complexity arising from the combination of diverse granularity of semantics, ranging from the local motion of body joints to high-level relationships across multiple people. To learn this multi-level characteristic of human action in an unsupervised manner, we propose a novel pretraining strategy along with a transformer-based model architecture named GL-Transformer++. Prior methods in unsupervised action recognition or unsupervised group activity recognition (GAR) have shown limitations, often focusing solely on capturing a partial scope of the action, such as the local movements of each individual or the broader context of the overall motion. To tackle this problem, we introduce a novel pretraining strategy named multi-interval pose displacement prediction (MPDP) that enables the model to learn the diverse extents of the action. In the architectural aspect, we incorporate the global and local attention (GLA) mechanism within the transformer blocks to learn local dynamics between joints, global context of each individual, as well as high-level interpersonal relationships in both spatial and temporal manner. In fact, the proposed method is a unified approach that demonstrates efficacy in both action recognition and GAR. Particularly, our method presents a new and strong baseline, surpassing the current SOTA GAR method by significant margins: 29.6% in Volleyball and 60.3% and 59.9% on the xsub and xset settings of the Mutual NTU dataset, respectively.

Accelerating large-scale multi-scalar multiplication in Zk-SNARK through exploiting its multilevel parallelism

In the context of zk-SNARK, MSM emerges as a major computational bottleneck, particularly due to its high computational and memory overhead. In this work, we exploit multiple levels of parallelism to effectively accelerate largescale MSM in zk-SNARKs. Firstly, a distributed parameter generation method is proposed in this paper to replace that of centralized method. Based on this methodology, the paper realizes a system with extraordinary scalability, capable of computing large-scale MSMs. Subsequently, the approach presented in this paper elevates the computation of Bellperson, the most prominent zero-knowledge proof system in real-world applications, from a single-node computation to a clustering mode, significantly enhancing its computational performance – a crucial advancement for practical applications. Finally, we implement a multi-level, fully parallelised MSM computing system by leveraging hierarchical sub-task partitioning and cross-node communication optimization, thereby thoroughly exploiting parallelism at diverse granularities. Experimental results show that in the cluster scenario, the proposed approach achieves acceleration ratios of approximately 3.60 and 6.50 times compared to the cutting-edge heterogeneous version Bellperson in dual-node and quad-node settings, respectively. On a single node, the proposed optimization approach achieves an acceleration ratio of 1.38 times compared to the current State-of-the-Art MSM calculation module of cuZK, outperforms the industry-popular Bellman by 186 times and the Bellperson by 1.96 times.

Fine-grained ship image classification and detection based on a vision transformer and multi-grain feature vector FPN model

ABSTRACT In naval and civilian domains, meticulous ship classification and detection are paramount. Nevertheless, predominant research has gravitated toward leveraging Convolutional Neural Network (CNN)-centered methodologies, often overlooking the diverse granularity inherent in ship samples. In our pursuit to holistically extract features from ship images across varying granularities, we present a transformative architecture: the Vision Transformer and Multi-Grain Feature Vector Feature Pyramid Network (ViT-MGFV-FPN). This model synergistically melds the merits of MGFV-FPN with an augmented Vision Transformer (ViT) for a comprehensive image feature extraction. To cater to the extraction of broader image features whilst sidestepping the innate quadratic complexity of traditional ViT, we unveil an enhanced version christened the Global Swin Transformer. Concurrently, the MGFV-FPN is orchestrated to harness the prowess of CNNs in distilling intricate ship attributes. Rigorous empirical evaluations underscore our model’s superiority in juxtaposition with extant CNN and transformer-based paradigms for nuanced ship categorization.

SLAFusion: Attention fusion based on SAX and LSTM for dangerous driving behavior detection

Dangerous driving behaviors are the main cause of most traffic accidents, and the detection of these behaviors is one of the extremely important researches in Intelligent Transportation System (ITS). Although some recent approaches consider dangerous driving behavior as a process containing multiple instantaneous points, which is ignored by traditional methods, they fail to take into account the loss of temporal information due to the reasonable division of instantaneous points. Therefore, this paper proposes a Symbolic Aggregate approXimation (SAX) and Long Short-Term Memory (LSTM)-based Attention Fusion method (SLAFusion) to detect dangerous driving behavior more accurately. This method evaluates various granularity of driving processes and explores the association between adjacent processes. First, a D-value Piecewise Aggregate Approximation (DPAA) is designed to model the change of vehicle state in mining driving processes with diverse granularity. Then, the features of multiple driving processes are fused to enhance each other to solve the problem of information loss due to segmentation of driving processes. Further, the auxiliary information is fused to incorporate the environmental features of the vehicle during driving. Finally, the experimental results on real datasets indicate that the proposed SLAFusion method improves the recall and F-score by 3.2% and 2.6% respectively in comparison with the existing benchmark methods.

Federated time persistency in intermittently powered IoT systems

The concept of intermittent computing has been recently proposed, which aims to remove batteries by adding small capacitors or directly powering computing systems from the energy harvesting source, retaining computation despite the intermittent nature of the supply. However, certain trivial functions of typical IoT platforms such as performing long-running computation and wireless transmission, or keeping track of time, pose a real challenge in a system with a small amount or total absence of energy storage. This paper proposes a novel framework for time persistency with diverse granularity, based on an hourglass strategy, which includes two small capacitors whose discharge rate is used to calculate the elapsed time from one power cycle to another. The proposed framework was implemented and tested in a real-life IoT system which is used to monitor the temperature of a motorised machine. The experimental results demonstrated an improvement of at least 73.4% in error rate of time measurements, and a minimum reduction of 48.3% in energy consumption against state-of-the-art approaches.

Document Summarization with VHTM: Variational Hierarchical Topic-Aware Mechanism

Automatic text summarization focuses on distilling summary information from texts. This research field has been considerably explored over the past decades because of its significant role in many natural language processing tasks; however, two challenging issues block its further development: (1) how to yield a summarization model embedding topic inference rather than extending with a pre-trained one and (2) how to merge the latent topics into diverse granularity levels. In this study, we propose a variational hierarchical model to holistically address both issues, dubbed VHTM. Different from the previous work assisted by a pre-trained single-grained topic model, VHTM is the first attempt to jointly accomplish summarization with topic inference via variational encoder-decoder and merge topics into multi-grained levels through topic embedding and attention. Comprehensive experiments validate the superior performance of VHTM compared with the baselines, accompanying with semantically consistent topics.

Managing Multigranular Linguistic Distribution Assessments in Large-Scale Multiattribute Group Decision Making

Linguistic large-scale group decision making (LGDM) problems are more and more common nowadays. In such problems a large group of decision makers are involved in the decision process and elicit linguistic information that are usually assessed in different linguistic scales with diverse granularity because of decision makers' distinct knowledge and background. To keep maximum information in initial stages of the linguistic LGDM problems, the use of multi-granular linguistic distribution assessments seems a suitable choice, however to manage such multigranular linguistic distribution assessments, it is necessary the development of a new linguistic computational approach. In this paper it is proposed a novel computational model based on the use of extended linguistic hierarchies, which not only can be used to operate with multi-granular linguistic distribution assessments, but also can provide interpretable linguistic results to decision makers. Based on this new linguistic computational model, an approach to linguistic large-scale multi-attribute group decision making is proposed and applied to a talent selection process in universities.

Localizing Wireless Sensors with Diverse Granularities in Wireless Sensor Networks

To obtain accurate location information of individual sensor nodes is of vital importance in wireless sensor networks (WSNs), especially for objective tracking applications. However, it is challenging to acquire fine-grained localization accuracy because of resource constraints of sensor nodes, unreliable wireless communication, and cost. Moreover, heterogeneous characteristics of both sensor nodes and applications make this problem even harder to solve. In this paper, we propose NLMR, a novel on-demand node localization technology based on multiresolution model. NLMR comprises three phases: (1) subregion classification, which categorizes regions into subregions with either uniform node deployment or nonuniform node deployment; (2) multiresolution model construction, which creates a multiresolution model that caters for diverse localization granularity; (3) node localization, which allows the control center to estimate the locations of sensor nodes in a centralized manner. Our analysis and simulation results demonstrate the performance of NLMR and verify that our scheme can provide diverse localization granularity with high probability.

Challenges and requirements of a control plane for elastic optical networks

Elastic optical networks have emerged as a promising technology for the efficient use of optical network resources. Its adaptable characteristics and adjustable data rate enable operators to meet the diverse granularity of their clients needs. In order to automate an elastic optical network operation, a control plane is required. Wavelength Switched Optical Networks (WSON) may already rely on a robust control plane which enables dynamic network management, provides prompt demand reply optimizing spectrum use, and implements important network features as survivability strategies, differentiated service, and grooming procedures. Due to its specific characteristics, elastic optical networks may not implement traditional WSON control plane solutions without further enhancement. Therefore, recent research efforts have been focusing on developing the control plane for this new technology, in most cases by proposing extensions to the currently available architectures. This paper describes a survey on the current ongoing research efforts to define elastic optical network control plane architecture. It identifies and classifies the most relevant proposals currently found in literature, and discusses how these propositions address the main requirements to design a control plane which enables automating the specific functions of an elastic optical network.

Statistical user model supported by R-Tree structure

This paper is about developing a group user model able to predict unknown features (attributes, preferences, or behaviors) of any interlocutor. Specifically, for systems where there are features that cannot be modeled by a domain expert within the human computer interaction. In such cases, statistical models are applied instead of stereotype user models. The time consumption of these models is high, and when a requisite of bounded response time is added most common solution involves summarizing knowledge. Summarization involves deleting knowledge from the knowledge base and probably losing accuracy in the medium-term. This proposal provides all the advantages of statistical user models and avoids knowledge loss by using an R-Tree structure and various search spaces (universes of users) of diverse granularity for solving inferences with enhanced success rates. Along with the formalization and evaluation of the approach, main advantages will be discussed, and a perspective for its future evolution is provided. In addition, this paper provides a framework to evaluate statistical user models and to enable performance comparison among different statistical user models.

Spatio-structural granularity of biological material entities.

BackgroundWith the continuously increasing demands on knowledge- and data-management that databases have to meet, ontologies and the theories of granularity they use become more and more important. Unfortunately, currently used theories and schemes of granularity unnecessarily limit the performance of ontologies due to two shortcomings: (i) they do not allow the integration of multiple granularity perspectives into one granularity framework; (ii) they are not applicable to cumulative-constitutively organized material entities, which cover most of the biomedical material entities.ResultsThe above mentioned shortcomings are responsible for the major inconsistencies in currently used spatio-structural granularity schemes. By using the Basic Formal Ontology (BFO) as a top-level ontology and Keet's general theory of granularity, a granularity framework is presented that is applicable to cumulative-constitutively organized material entities. It provides a scheme for granulating complex material entities into their constitutive and regional parts by integrating various compositional and spatial granularity perspectives. Within a scale dependent resolution perspective, it even allows distinguishing different types of representations of the same material entity. Within other scale dependent perspectives, which are based on specific types of measurements (e.g. weight, volume, etc.), the possibility of organizing instances of material entities independent of their parthood relations and only according to increasing measures is provided as well. All granularity perspectives are connected to one another through overcrossing granularity levels, together forming an integrated whole that uses the compositional object perspective as an integrating backbone. This granularity framework allows to consistently assign structural granularity values to all different types of material entities.ConclusionsThe here presented framework provides a spatio-structural granularity framework for all domain reference ontologies that model cumulative-constitutively organized material entities. With its multi-perspectives approach it allows querying an ontology stored in a database at one's own desired different levels of detail: The contents of a database can be organized according to diverse granularity perspectives, which in their turn provide different views on its content (i.e. data, knowledge), each organized into different levels of detail.

Semantic and pedagogic interoperability mechanisms in the ARIADNE educational repository

This paper reports on the principles underlying the semantic and pedagogic interoperability mechanisms built in the European Knowledge Pool System, developed by the European research project ARIADNE. This system, which is the central feature of ARIADNE, consists in a distributed repository of pedagogical documents (or learning objects) of diverse granularity, origin, content, type, language, etc., which are stored in view of their use (and reuse) in telematics-based training or teaching curricula. The learning objects are indexed, usually by faculty staff, according to the ARIADNE metadata set. The principles embodied in the indexation tool, which interacts directly with the repository,stem from a few theoretical ideas but foremost from empirical, pragmatic considerations, suggested by the context of actual use. They tentatively address the stringent demands for semantic and pedagogic interoperability implied by a context of rather wide cultural and linguistic diversity, as well as those stemming from the very nature of the domain application itself: education and training. Possible extensions to the educational metadata scheme developed by ARIADNE on these basis, may accommodate corporate training/information needs. These extensions are briefly discussed as a mean for enhancing 'semantic' interoperability between different (kinds of) corporations. Finally, the architecture of the ARIADNE system, which heavily relies on this educational metadata system, is briefly reviewed.