Fine Granularity Research Articles

Fine-grained selective similarity integration for drug-target interaction prediction.

The discovery of drug-target interactions (DTIs) is a pivotal process in pharmaceutical development. Computational approaches are a promising and efficient alternative to tedious and costly wet-lab experiments for predicting novel DTIs from numerous candidates. Recently, with the availability of abundant heterogeneous biological information from diverse data sources, computational methods have been able to leverage multiple drug and target similarities to boost the performance of DTI prediction. Similarity integration is an effective and flexible strategy to extract crucial information across complementary similarity views, providing a compressed input for any similarity-based DTI prediction model. However, existing similarity integration methods filter and fuse similarities from a global perspective, neglecting the utility of similarity views for each drug and target. In this study, we propose a Fine-Grained Selective similarity integration approach, called FGS, which employs a local interaction consistency-based weight matrix to capture and exploit the importance of similarities at a finer granularity in both similarity selection and combination steps. We evaluate FGS on five DTI prediction datasets under various prediction settings. Experimental results show that our method not only outperforms similarity integration competitors with comparable computational costs, but also achieves better prediction performance than state-of-the-art DTI prediction approaches by collaborating with conventional base models. Furthermore, case studies on the analysis of similarity weights and on the verification of novel predictions confirm the practical ability of FGS.

2-D fragmentation-aware RMSA based on the relative flexibility of advance reservation requests in elastic optical networks

With the advantages of fine granularity spectrum allocation (SA) and large capacity, elastic optical networks (EONs) are considered as the most promising development direction of optical network in the future. The emergence of two-dimensional (2-D) fragmentation during dynamic SA of EONs not only reduces the availability of resources but also increases the blocking probability of requests. In addition, the advance reservation (AR) request has the characteristics of resource reservation and time flexibility, which further complicates the fragmentation issue. To solve the above issues, in this paper, a fragmentation-aware routing, modulation and spectrum assignment (RMSA) algorithm for AR requests is proposed. First, delayed resource allocation is developed for AR requests. Second, a resource allocation sorting strategy based on the relative flexibility of requests is proposed. The strategy makes full use of the time characteristics of AR request to determine the allocation order of requests. In addition, from the perspective of the impact on adjacent resources, a 2-D path resource model is introduced to evaluate the impact of 2-D fragmentation from the perspective of resource matching degree and resource connectivity. Moreover, simulation results show that the proposed algorithm is better than the benchmark algorithm in terms of blocking probability, resource utilization and fragmentation rate.

Urban land use transitions: Examining change over 19 years using sequence analysis. The case of South-East Queensland, Australia

The land use mosaic that characterises our urban environments is complex and subject to regular and on-going change and transition. Land use change takes place as cities seek to meet ever evolving population, economic, social, and environmental objectives. However, our empirical capacity to map, measure and monitor the geographical shifts in land use at a fine spatial granularity and how these aggregate across the urban environment remain very limited. In this paper, we draw on parcel level land use data for a large metropolitan region in Australia for a 19-year period and employ sequence analysis to delineate the location and timing of shifts in land use. Results reveal both similarities between jurisdictional regions alongside the unique land use transitions that go some way to highlight context specific mechanisms. This study demonstrates the utility of our empirical approach in its capacity to inform regional development strategies through revealing the type, timing and location of land use change in relation to land use policy and planning goals.

TTL: transformer-based two-phase transfer learning for cross-lingual news event detection

Today, we have access to a vast data amount, especially on the internet. Online news agencies play a vital role in this data generation, but most of their data is unstructured, requiring an enormous effort to extract important information. Thus, automated intelligent event detection mechanisms are invaluable to the community. In this research, we focus on identifying event details at the sentence and token levels from news articles, considering their fine granularity. Previous research has proposed various approaches ranging from traditional machine learning to deep learning, targeting event detection at these levels. Among these approaches, transformer-based approaches performed best, utilising transformers’ transferability and context awareness, and achieved state-of-the-art results. However, they considered sentence and token level tasks as separate tasks even though their interconnections can be utilised for mutual task improvements. To fill this gap, we propose a novel learning strategy named Two-phase Transfer Learning (TTL) based on transformers, which allows the model to utilise the knowledge from a task at a particular data granularity for another task at different data granularity, and evaluate its performance in sentence and token level event detection. Also, we empirically evaluate how the event detection performance can be improved for different languages (high- and low-resource), involving monolingual and multilingual pre-trained transformers and language-based learning strategies along with the proposed learning strategy. Our findings mainly indicate the effectiveness of multilingual models in low-resource language event detection. Also, TTL can further improve model performance, depending on the involved tasks’ learning order and their relatedness concerning final predictions.

A spatiotemporal graph wavelet neural network for traffic flow prediction

The traffic flow prediction is fast becoming a key instrument in the transportation system, which has achieved impressive performance for traffic management. The graph neural network plays a critical role in the development of the traffic network management. However, it is worthwhile mentioning that the complexity of road networks and traffic conditions makes it unable to obtain sufficient spatiotemporal information. In view of capturing precise environment characteristics, the context could have a precise effect on the prediction results while previous methods rarely took this into account. Besides, the nonlinear characteristics of the graph neural network are hard to quantify with fine granularity and to eliminate overfitting. To stack these challenges, in this paper, we present a spatiotemporal graph wavelet neural network to improve the ability of representations. Specifically, we introduce the wavelet transforms into the deep learning model according to the strong nonlinear optimization ability. Furthermore, we dig the location and time patterns to evaluate the temporal dependence and the spatial proximity correlation. In addition, we introduce a historical context attention mechanism giving fine-grained historical context grade evaluation to ease the phenomenon of over-smoothing. The experimental results on real-world datasets show that our work gets considerable results compared with the baseline and start-of-the-art models. Moreover, our work has better learning performance by employing the connection and interaction of graphs.

CADEN: Compression-Assisted Adaptive Encoding to Improve Lifetime of Encrypted Nonvolatile Main Memories

Nonvolatile memories are widely considered as a potential replacement of DRAM in main memory due to their high density and low-leakage power consumption. The data in these memories are stored in encrypted form to protect them from data stealing. However, the encryption techniques, on account of their diffusion property impose high randomization in the generated encrypted data. It leads to enormous bit-flips in the memory cells, leading to their early wear out. In this letter, we propose a technique called CADEN that combines our compression technique COMF with an adaptive encoding scheme. The compressed data generated by COMF is encoded in finer granularity using an adaptive encoding approach at low storage overhead. Experimental evaluation shows that CADEN reduces bit-flips and improves lifetime in PCM main memory compared to baseline and existing techniques. In particular, CADEN shows 52% and 57% reduction in bit-flips and energy consumption and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$2.31\times $ </tex-math></inline-formula> improvement in lifetime over baseline.

Scaling Identifiers and their Metadata to Gigascale: An Architecture to Tackle the Challenges of Volume and Variety

Persistent identifiers are applied to an ever-increasing variety of research objects, including software, samples, models, people, instruments, grants, and projects, and there is a growing need to apply identifiers at a finer and finer granularity. Unfortunately, the systems developed over two decades ago to manage identifiers and the metadata describing the identified objects no longer scale. Communities working with physical samples have grappled with these three challenges of the increasing volume, variety, and variability of identified objects for many years. To address this dual challenge, the IGSN 2040 project explored how metadata and catalogues for physical samples could be shared at the scale of billions of samples across an ever-growing variety of users and disciplines. In this paper, we focus on how we scale identifiers and their describing metadata to billions of objects and who the actors involved with this system are. Our analysis of these requirements resulted in the definition of a minimum viable product and the design of an architecture that not only addresses the challenges of increasing volume and variety but, more importantly, is easy to implement because it reuses commonly used Web components. Our solution is based on a Web architectural model that utilises <a href="http://Schema.org">Schema.org</a>, JSON-LD, and sitemaps. Applying these commonly used architectural patterns on the internet allows us to not only handle increasing variety but also enable better compliance with the FAIR Guiding Principles.

Approachability and Credibility of Virtual Character Faces: The Role of the Horizontal Viewing Angle.

The present work explores how the horizontal viewing angle of a virtual character's face influences perceptions of credibility and approachability. When encountering virtual characters, people rely both on credibility and approachability judgments to form a first impression of the depicted virtual character. Research shows that certain perceptions are preferred either on frontal or tilted faces, but not how approachability or credibility judgments relate to horizontal viewing angles in finer granularity between 0° and 45°. 52 participants performed a two-alternative forced choice (2AFC) task rating 240 pairwise comparisons of 20 virtual character faces shown in four horizontal viewing angles (0°, 15°, 30°, and 45°) on approachability and credibility. They also rated scales on individual differences based on the BIS-BAS framework (behavioral inhibition system, drive, and reward responsiveness), self-esteem, and personality traits (neuroticism, loneliness). Both approachability and credibility were negatively related to the horizontal viewing angle, but the negative relationship was less pronounced for approachability. Notably, 15° tilted faces were associated with higher approachability than frontal faces by people scoring high in reward responsiveness, drive, and self-esteem, and scoring low in neuroticism and loneliness. Our findings highlight the conditions under which showing a virtual character's face is preferred in a horizontally 15° tilted over a frontal position. The differential impact of the horizontal viewing angle on approachability and credibility should be considered when displaying virtual character faces.

Mining of EHR for interface terminology concepts for annotating EHRs of COVID patients

BackgroundTwo years into the COVID-19 pandemic and with more than five million deaths worldwide, the healthcare establishment continues to struggle with every new wave of the pandemic resulting from a new coronavirus variant. Research has demonstrated that there are variations in the symptoms, and even in the order of symptom presentations, in COVID-19 patients infected by different SARS-CoV-2 variants (e.g., Alpha and Omicron). Textual data in the form of admission notes and physician notes in the Electronic Health Records (EHRs) is rich in information regarding the symptoms and their orders of presentation. Unstructured EHR data is often underutilized in research due to the lack of annotations that enable automatic extraction of useful information from the available extensive volumes of textual data.MethodsWe present the design of a COVID Interface Terminology (CIT), not just a generic COVID-19 terminology, but one serving a specific purpose of enabling automatic annotation of EHRs of COVID-19 patients. CIT was constructed by integrating existing COVID-related ontologies and mining additional fine granularity concepts from clinical notes. The iterative mining approach utilized the techniques of 'anchoring' and 'concatenation' to identify potential fine granularity concepts to be added to the CIT. We also tested the generalizability of our approach on a hold-out dataset and compared the annotation coverage to the coverage obtained for the dataset used to build the CIT.ResultsOur experiments demonstrate that this approach results in higher annotation coverage compared to existing ontologies such as SNOMED CT and Coronavirus Infectious Disease Ontology (CIDO). The final version of CIT achieved about 20% more coverage than SNOMED CT and 50% more coverage than CIDO. In the future, the concepts mined and added into CIT could be used as training data for machine learning models for mining even more concepts into CIT and further increasing the annotation coverage.ConclusionIn this paper, we demonstrated the construction of a COVID interface terminology that can be utilized for automatically annotating EHRs of COVID-19 patients. The techniques presented can identify frequently documented fine granularity concepts that are missing in other ontologies thereby increasing the annotation coverage.

Defining the distance between diseases using SNOMED CT embeddings

Characterizing disease relationships is essential to biomedical research to understand disease etiology and improve clinical decision-making. Measurements of distance between disease pairs enable valuable research tasks, such as subgrouping patients and identifying common time courses of disease onset. Distance metrics developed in prior work focused on smaller, targeted disease sets. Distance metrics covering all diseases have not yet been defined, which limits the applications to a broader disease spectrum. Our current study defines disease distances for all disease pairs within the International Classification of Diseases, version 10 (ICD-10), the diagnostic classification system universally used in electronic health records. Our proposed distance is computed based on a biomedical ontology, SNOMED CT (Systemized Nomenclature of Medicine, Clinical Terms), which can also be viewed as a structured knowledge graph. We compared the knowledge graph-based metric to three other distance metrics based on the hierarchical structure of ICD, clinical comorbidity, and genetic correlation, to evaluate how each may capture similar or unique aspects of disease relationships. We show that our knowledge graph-based distance metric captures known phenotypic, clinical, and molecular characteristics at a finer granularity than the other three. With the continued growth of using electronic health records data for research, we believe that our distance metric will play an important role in subgrouping patients for precision health, and enabling individualized disease prevention and treatments.

Attention-based deep meta-transfer learning for few-shot fine-grained fault diagnosis

Deep learning-based fault diagnosis methods have made tremendous progress in recent years; however, most of these methods are coarse grained and data demanding that cannot find the root causes of mechanical system failures at a finer granularity with limited fault data. Therefore, in this study, we first investigate the few-shot fine-grained fault diagnosis (FSFGFD) problem, with the aim of identifying novel fine-grained faults under different working conditions using only few samples from each class. To address the difficulties of fine-grained fault feature extraction and poor model generalization to unseen few-shot faults in FSFGFD tasks, a novel attention-based deep meta-transfer learning (ADMTL) method is proposed. First, the failure modes under different working conditions are considered as fine-grained faults, and their raw signals are transformed into time–frequency images. Based on this, an attention mechanism is introduced to guide the feature extractor of the ADMTL on what information to learn. The ADMTL then follows a three-stage learning process of pre-training, meta-transfer, and meta-adaptation to achieve fast adaptation to new fine-grained faults using a priori knowledge gained from known faults. Furthermore, a parameter modulation strategy is employed to adaptively update the pre-trained network during the meta-transfer process. The comprehensive experimental results of three case studies demonstrate the superiority of our method over state-of-the-art methods. The proposed method achieves excellent performance with an average accuracy of 99.08%, 95.86%, and 77.74% for FSFGFD tasks when performing meta-transfer within the same machine and between different machines, respectively.

Routing and Timeslot Scheduling for SPN Fine-Granularity Slices

The integration of 5G and vertical industries promotes the development of the energy Ethernet while putting forward fine granularity, flexibility, high reliability, and deterministic low-latency service requirements for the smart grid and the ubiquitous power Internet of Things (UPIoT). As the bearer architecture supporting the next-generation optical transmission network, the Slicing Packet Network (SPN) slice granularity decreases from 5 Gbps to 10 Mbps fine granularity and the frame period of 5 Gbps large-granularity slices is short, so the non-deterministic delay caused by timeslot conflicts has a negligible impact on the end-to-end delay, and the timeslot scheduling is unnecessary. However, due to the reduction in timeslot granularity and the change in frame structure in 10 Mbps slices, the scheduling of conflicting timeslots and the complex device computing management problems need to be solved urgently. In this paper, we establish a model of routing embedded timeslot scheduling for the routing of fine-granularity slices and timeslot scheduling problems in SPN-based FlexE interfaces, for which we propose a deterministic timeslot allocation mechanism supporting end-to-end low-latency transmission. According to the timeslot symmetry, the mechanism can reduce the space of feasible solutions through ant colony optimization and unidirectional neighborhood search (ACO-UNS), so as to efficiently solve the scheduling of conflicting timeslots and provide end-to-end delay guarantee for delay-sensitive services. Finally, we make a comparison between the ACO-UNS algorithm and the timeslot random dispatching algorithm (ACO-RD); the results show that, relative to the ACO-RD, the reduction in the proposed ACO-UNS is 98.721% for the end-to-end delay of fine-granularity slices.

Defect identification in adhesive structures using multi-Feature fusion convolutional neural network

The interface-debonding defects of adhesive bonding structures may cause a reduction in bonding strength, which in turn affects the bonding quality of adhesive bonding samples. Hence, defect recognition in adhesive bonding structures is particularly important. In this study, a terahertz (THz) wave was used to analyze bonded structure samples, and a multi-feature fusion convolutional neural network (CNN) was used to identify the defect waveforms. The pooling method of the squeeze-and-excitation (SE) attention mechanism was optimized, defect feature weights were adaptively assigned, and feature fusion was conducted using automatic label net-works to segment the THz waveforms in the adhesive bonding area with fine granularity waveforms as an input to the multi-channel CNN. The results revealed that the speed of the THz waveform labeling with the automatic labeling network was 10 times higher than that with traditional methods, and the defect-recognition accuracy of the defect-recognition network constructed in this study was up to 99.28%. The F1-score was 99.73%, and the lowest pre-embedded defect recognition error rate of the generalization experiment samples was 0.27%.

KEMIM: Knowledge-Enhanced User Multi-Interest Modeling for Recommender Systems

Researchers typically leverage side information, such as social networks or the knowledge graph, to overcome the sparsity and cold start problem in collaborative filtering. To tackle the limitations of existing user interest modeling, we propose a knowledge-enhanced user multi-interest modeling for recommender systems (KEMIM). First, we utilize the user-item historical interaction as the knowledge graph’s head entity to create a user’s explicit interests and leverage the relationship path to expand the user’s potential interests through connections in the knowledge graph. Second, considering the diversity of a user’s interests, we adopt an attention mechanism to learn the user’s attention to each historical interaction and each potential interest. Third, we combine the user’s attribute features with interests to solve the cold start problem effectively. With the knowledge graph’s structural data, KEMIM could describe the features of users at a fine granularity and provide explainable recommendation results to users. In this study, we conduct an in-depth empirical evaluation across three open datasets for two different recommendation tasks: Click-Through rate (CTR) prediction and Top-K recommendation. The experimental findings demonstrate that KEMIM outperforms several state-of-the-art baselines.

Distributional Synthetic Controls

The method of synthetic controls is a fundamental tool for evaluating causal effects of policy changes in settings with observational data. In many settings where it is applicable, researchers want to identify causal effects of policy changes on a treated unit at an aggregate level while having access to data at a finer granularity. This article proposes an extension of the synthetic controls estimator that takes advantage of this additional structure and provides nonparametric estimates of the heterogeneity within the aggregate unit. The idea is to replicate the quantile function associated with the treated unit by a weighted average of quantile functions of the control units. This estimator relies on the same mathematical theory as the changes‐in‐changes estimator and can be applied in both repeated cross‐sections and panel data with as little as a single pre‐treatment period. It also provides a unique counterfactual quantile function for any type of distribution.

Dynamic Resource Allocation Using an Adaptive Multi-Objective Teaching-Learning Based Optimization Algorithm in Cloud

Resource allocation is a non-polynomial complete problem in the cloud data center that selects the proper resources to execute many fine computational granularity tasks. Customer requirements and capacity of applications change frequently. To bridge the gap between frequently changing customer requirement and available infrastructure for the services, we propose a dynamic resource allocation strategy using an adaptive multi-objective teaching-learning based optimization (AMO-TLBO) algorithm in Cloud computing. To improve the exploration and exploitation capacities, AMO-TLBO introduces the concept of number of teachers, adaptive teaching factor, tutorial training and self-motivated learning. Moreover, a grid-based approach to adaptively assess the non-dominated solutions maintained in an external archive is used. The objectives of AMO-TLBO include minimizing makespan, cost and maximizing utilization using well-balanced load across virtual machines. The evaluation results show that the proposed algorithm outperforms TLBO, MOPSO and NSGA-II algorithms in terms of different performance metrics.

Progressive Contextual Aggregation Empowered by Pixel-wise Confidence Scoring for Image Inpainting.

Image inpainting methods leverage the similarity of adjacent pixels to create alternative content. However, as the invisible region becomes larger, the pixels completed in the deeper hole are difficult to infer from the surrounding pixel signal, which is more prone to visual artifacts. To help fill this void, we adopt an alternative progressive hole-filling scheme that hierarchically fills the corrupted region in the feature and image spaces. This technique allows us to utilize reliable contextual information of the surrounding pixels, even for large hole samples, and then gradually complete the details as the resolution increases. For a more realistic representation of the completed region, we devise a pixel-wise dense detector. By distinguishing each pixel as whether it is a masked region or not, and passing the gradient to all resolutions, the generator further enhances the potential quality of the compositing. Furthermore, the completed images at different resolutions are then merged using a proposed structure transfer module (STM) that incorporates fine-grained local and coarse-grained global interactions. In this new mechanism, each completed image at the different resolutions attends its closest composition at fine granularity adjacent image and thus can capture the global continuity by interacting both short- and long-range dependencies. By comparing our solutions qualitatively and quantitatively with state-of-the-art methods, we conclude that our model exhibits a significantly improved visual quality, even in the case of large holes.

Periodic Physical Activity Information Segmentation, Counting and Recognition From Video

The research on complex human body motion including sports and workout activity recognition is a major challenge and long-lasting problem for the computer vision community. Recent development in deep learning algorithms to track people’s workout activity characteristics based on video sensors can be used to infer the human body pose for further analysis. Specifically, tracking complex body movements while performing multi-pose physical exercise helps individuals provide fine granularity feedback including activity repetition counting and activity recognition. Therefore, this research proposes a system that provides a repetition counter and activity recognition of physical exercise from video frames (extracted 3D human skeleton using VIBE) based on the deep semantic features and repetitive segmentation algorithm. The proposed system locates both ends of the activity’s action and segments the activity into multiple unit actions which improves activity recognition, time intervals, # of sets, and other quantitative values of activity. The proposed system is evaluated on the physical activities dataset named “NOL-18 Exercise” through extensive experiments. The proposed system results show that the accuracy of the repetitive action segmentation is 96.27% with 0.23% time error, and action recognition reaches 99.06%. The system can be employed to fitness or rehabilitation centers and used for treating patients.

A hybrid model for high spatial and temporal resolution population distribution prediction

ABSTRACT The accurate prediction of population distribution is crucial for numerous applications, from urban planning to epidemiological modelling. Using one-week data collected from open and multiple sources, including telecommunication activity, weather, point of interest, buildings, roads, and land use in Milan, Italy, we develop a hybrid method combining cellular automata (CA) and long short-term memory (LSTM) to predict population distribution with fine temporal and spatial granularity. Specifically, the convolutional autoencoder and LightGBM are applied to identify missing building types based on the pedestrian shed. The LSTM learns the transition rules of CA and Shapley additive explanations value is used for variable importance analysis. Results demonstrate that the combination of convolutional autoencoder and LightGBM is effective in building type prediction. The proposed model for population distribution prediction outperforms LSTM, the combination of CA and neural network, and the combination of CA and LightGBM by at least 5–10%. A variable importance analysis reveals that temporal variables are the most significant for prediction, followed by spatial and natural variables. The order of hour, housing-related variables, and types of precipitation are the most important variables in each category.

HDGFI: Hierarchical Dual-Level Graph Feature Interaction Model for Personalized Recommendation.

Under the background of information overload, the recommendation system has attracted wide attention as one of the most important means for this problem. Feature interaction considers not only the impact of each feature but also the combination of two or more features, which has become an important research field in recommendation systems. There are two essential problems in current feature interaction research. One is that not all feature interactions can generate positive gains, and some may lead to an increase in noise. The other is that the process of feature interactions is implicit and uninterpretable. In this paper, a Hierarchical Dual-level Graph Feature Interaction (HDGFI) model is proposed to solve these problems in the recommendation system. The model regards features as nodes and edges as interactions between features in the graph structure. Interaction noise is filtered by beneficial interaction selection based on a hierarchical edge selection module. At the same time, the importance of interaction between nodes is modeled in two perspectives in order to learn the representation of feature nodes at a finer granularity. Experimental results show that the proposed HDGFI model has higher accuracy than the existing models.