Scale Datasets Research Articles

TriMLP : A Foundational MLP-Like Architecture for Sequential Recommendation

In this work, we present TriMLP as a foundational MLP-like architecture for the sequential recommendation, simultaneously achieving computational efficiency and promising performance. First, we empirically study the incompatibility between existing purely MLP-based models and sequential recommendation, that the inherent fully-connective structure endows historical user–item interactions (referred as tokens) with unrestricted communications and overlooks the essential chronological order in sequences. Then, we propose the MLP-based Triangular Mixer to establish ordered contact among tokens and excavate the primary sequential modeling capability under the standard auto-regressive training fashion. It contains (1) a global mixing layer that drops the lower-triangle neurons in MLP to block the anti-chronological connections from future tokens and (2) a local mixing layer that further disables specific upper-triangle neurons to split the sequence as multiple independent sessions. The mixer serially alternates these two layers to support fine-grained preferences modeling, where the global one focuses on the long-range dependency in the whole sequence, and the local one calls for the short-term patterns in sessions. Experimental results on 12 datasets of different scales from 4 benchmarks elucidate that TriMLP consistently attains favorable accuracy/efficiency tradeoff over all validated datasets, where the average performance boost against several state-of-the-art baselines achieves up to 14.88%, and the maximum reduction of inference time reaches 23.73%. The intriguing properties render TriMLP a strong contender to the well-established RNN-, CNN-, and Transformer-based sequential recommenders. Code is available at https://github.com/jiangyiheng1/TriMLP .

Intelligent Fault Diagnosis in Industrial Machinery: Leveraging AI with LSTM Autoencoder for Enhanced Fault Detection

Machinery Fault Detection (MFD) is an important process in contemporary industrial systems, where it predicts possible physical failures before they lead to a serious problem. This uses multiple technologies to monitor machine statuses (algorithms, data gathering systems and sensors) Using a servo-motor driven actuator for deployment, the Locking Mechanism is pre-assembled into an OEM ATE and will enable predictive failure mode identification (via monitoring and warnings of operational parameters i.e., vibration, temperature or auditory signals in-built to MFD systems) leading to Prophylactic maintenance before critical bottlenecks can occur. The dataset we used in our study was collected from Kaggle and it is called the SpectraQuest Machinery Fault Simulator (MFS) Alignment-Balance-Vibration (ABVT). We used LSTM Autoencoder, KNN, SVM and DNN to analyzed the data. Our LSTM Autoencoder model was very accurate and achieved a precision, recall, accuracy and F-score of 99%. We worked on very large scale datasets. It will help the system detect faults and predict their evolution over time, so you save maintenance costs and increase production in your factory. More research on the practical efficiency of these models in real-time across different industrial settings can create a path towards improved and scalable MFD solutions.

Main challenges on the curation of large scale datasets for pancreas segmentation using deep learning in multi-phase CT scans: Focus on cardinality, manual refinement, and annotation quality

Accurate segmentation of the pancreas in computed tomography (CT) holds paramount importance in diagnostics, surgical planning, and interventions. Recent studies have proposed supervised deep-learning models for segmentation, but their efficacy relies on the quality and quantity of the training data. Most of such works employed small-scale public datasets, without proving the efficacy of generalization to external datasets. This study explored the optimization of pancreas segmentation accuracy by pinpointing the ideal dataset size, understanding resource implications, examining manual refinement impact, and assessing the influence of anatomical subregions. We present the AIMS-1300 dataset encompassing 1,300 CT scans. Its manual annotation by medical experts required 938 h. A 2.5D UNet was implemented to assess the impact of training sample size on segmentation accuracy by partitioning the original AIMS-1300 dataset into 11 smaller subsets of progressively increasing numerosity. The findings revealed that training sets exceeding 440 CTs did not lead to better segmentation performance. In contrast, nnU-Net and UNet with Attention Gate reached a plateau for 585 CTs. Tests on generalization on the publicly available AMOS-CT dataset confirmed this outcome. As the size of the partition of the AIMS-1300 training set increases, the number of error slices decreases, reaching a minimum with 730 and 440 CTs, for AIMS-1300 and AMOS-CT datasets, respectively. Segmentation metrics on the AIMS-1300 and AMOS-CT datasets improved more on the head than the body and tail of the pancreas as the dataset size increased. By carefully considering the task and the characteristics of the available data, researchers can develop deep learning models without sacrificing performance even with limited data. This could accelerate developing and deploying artificial intelligence tools for pancreas surgery and other surgical data science applications.

AFC-Unet: Attention-fused full-scale CNN-transformer unet for medical image segmentation

In the field of medical image segmentation, although U-Net has achieved significant achievements, it still exposes some inherent disadvantages when dealing with complex anatomical structures and small targets, such as inaccurate target localization, blurry edges, and insufficient integration of contextual information. To address these challenges, this study proposes the Attention-Fused Full-Scale CNN-Transformer Unet (AFC-Unet), aiming to effectively overcome the limitations of traditional U-Net through multi-scale feature fusion, attention mechanisms, and CNN-Transformer hybrid modules. Specifically, we adopt an encoder–decoder architecture, incorporating full-scale feature block fusion and pyramid sampling modules to enhance the model’s ability to recognize fine to overall structural features by integrating cross-level multi-scale features. We propose the Multi-feature Fusion Attention Gates (MFAG) module, which focuses on and highlights discriminative information of potential lesions and key anatomical boundaries, effectively suppressing irrelevant background interference. We design a module Convolutional Hybrid Attention Transformer (CHAT) that integrates CNN and Transformer to address the shortcomings of traditional single models in handling long-range dependencies and global context understanding. Experimental results on three datasets of different scales demonstrate that the model’s segmentation performance for medical images surpasses state-of-the-art models, showcasing high generalization ability.

Fast Multiview Clustering by Optimal Graph Mining.

Multiview clustering (MVC) aims to exploit heterogeneous information from different sources and was extensively investigated in the past decade. However, far less attention has been paid to handling large-scale multiview data. In this brief, we fill this gap and propose a fast multiview clustering by an optimal graph mining model to handle large-scale data. We mine a consistent clustering structure from landmark-based graphs of different views, from which the optimal graph based on the one-hot encoding of cluster labels is recovered. Our model is parameter-free, so intractable hyperparameter tuning is avoided. An efficient algorithm of linear complexity to the number of samples is developed to solve the optimization problems. Extensive experiments on real-world datasets of various scales demonstrate the superiority of our proposal.

Real-space visualization of a defect-mediated charge density wave transition

We study the coupled charge density wave (CDW) and insulator-to-metal transitions in the 2D quantum material 1T-TaS2. By applying in situ cryogenic 4D scanning transmission electron microscopy with in situ electrical resistance measurements, we directly visualize the CDW transition and establish that the transition is mediated by basal dislocations (stacking solitons). We find that dislocations can both nucleate and pin the transition and locally alter the transition temperature Tc by nearly ~75 K. This finding was enabled by the application of unsupervised machine learning to cluster five-dimensional, terabyte scale datasets, which demonstrate a one-to-one correlation between resistance-a global property-and local CDW domain-dislocation dynamics, thereby linking the material microstructure to device properties. This work represents a major step toward defect-engineering of quantum materials, which will become increasingly important as we aim to utilize such materials in real devices.

Energy-saving scheduling strategy for variable-speed flexible job-shop problem considering operation-dependent energy consumption

The rapid growth of industrial production signifies its CO2 footprint and climate impact, highlighting the demand for efficient and energy-saving production scheduling. Generally, a single production job comprises multiple operations on various machines. Existing studies estimate job energy consumption from a broader perspective, often disregarding the detailed scheduling of operations. This simplification may result in inaccurate evaluations of energy consumption and savings, as energy consumption can differ significantly among operations, with each requiring specific levels of energy on various machines. Proposing a new scheduling framework to replace the classical strategy is challenging yet necessary due to these variations. This study aims to investigate the multiobjective flexible job-shop scheduling problem by considering operation-dependent energy consumption and to improve the accuracy and thoroughness of the energy-consumption evaluation framework. First, a core metamodel for energy assessment is established and a mathematical model is introduced to minimize the manufacturing span and total energy consumption. Subsequently, a dynamic diffusion weight adjustment for a multiobjective evolutionary algorithm based on decomposition (MOEA/D-DDWA) is proposed. Operation-dependent energy consumption and processing quality are discussed, and a superior unified energy-saving strategy is designed to facilitate the selection of the processing-speed range, balancing energy consumption and saving. Finally, numerical experiments are conducted based on datasets of various scales, demonstrating that the core metamodel for energy assessment can achieve significant energy savings. Compared with other classical multiobjective optimization algorithms, the MOEA/D-DDWA is more energy efficient.

Multi-Objective Evolutionary Neural Architecture Search with Weight-Sharing Supernet

Deep neural networks have played a crucial role in the field of deep learning, achieving significant success in practical applications. The architecture of neural networks is key to their performance. In the past few years, these architectures have been manually designed by experts with rich domain knowledge. Additionally, the optimal neural network architecture can vary depending on specific tasks and data distributions. Neural Architecture Search (NAS) is a class of techniques aimed at automatically searching for and designing neural network architectures according to the given tasks and data. Specifically, evolutionary-computation-based NAS methods are known for their strong global search capability and have aroused widespread interest in recent years. Although evolutionary-computation-based NAS has achieved success in a wide range of research and applications, it still faces bottlenecks in training and evaluating a large number of individuals during optimization. In this study, we first devise a multi-objective evolutionary NAS framework based on a weight-sharing supernet to improve the search efficiency of traditional evolutionary-computation-based NAS. This framework combines the population optimization characteristic of evolutionary algorithms with the weight-sharing ideas in one-shot models. We then design a bi-population MOEA/D algorithm based on the proposed framework to effectively solve the NAS problem. By constructing two sub-populations with different optimization objectives, the algorithm can effectively explore network architectures of various sizes in complex search spaces. An inter-population communication mechanism further enhances the algorithm’s exploratory capability, enabling it to find network architectures with uniform distribution and high diversity. Finally, we conduct performance comparison experiments on image classification datasets of different scales and complexities. Experimental results demonstrate the effectiveness of the proposed multi-objective evolutionary NAS framework and the practicality and transferability of the introduced bi-population MOEA/D-based NAS method compared to existing state-of-the-art NAS methods.

A novel data-driven IBA-ELM model for SOH/SOC estimation of lithium-ion batteries

Lithium-ion batteries (Libs) have become the best power source for electric vehicles due to their high energy density and long cycle life. The state estimation is directly related to the safe and efficient use of Libs. In our previous research, a Bat Algorithm-Extreme Learning Machine (BA-ELM) model was constructed and promising results were achieved for state of health (SOH) estimation. To increase the engineering application value of BA-ELM for SOH/state of charge (SOC) estimation, optimization mechanism of improving the global exploration and local exploitation capabilities of the model was studied. First, three improved methods including a conversion mechanism of Artificial Bee Colony algorithm employing bees and scout bees, a mechanism for adaptive adjustment of the search frequency of bats according to the Euclidean distance difference, and an adaptive inertia weight coefficient method were proposed to construct a novel Improved Bat Algorithm (IBA)-ELM model. Second, a battery life degradation test platform was built to obtain the degradation dataset for Libs. The variation rules of nine battery variables and cycles were revealed. Four high-quality SOH-related features were extracted. Thus, the SOH estimation performance of six models was evaluated. Third, a Libs charge–discharge dataset under driving conditions was collected and processed. Four models were trained based on 72,048 samples from four conditions. Further, SOC estimation tests were conducted on 35,071 samples from completely different conditions. The results show that the root mean square error (RMSE) and mean absolute error (MAE) of SOH estimation by IBA-ELM decreased by 21.93 % and 23.79 %, respectively, compared to those by BA-ELM. The RMSE and MAE of SOC estimation decreased by more than 10 %. Error, radar, and violin plots demonstrate that the IBA-ELM could accurately estimate SOH/SOC in larger scale datasets, simultaneously exhibiting better generalization performance.

Rule learning by modularity

In this paper, we present a modular methodology that combines state-of-the-art methods in (stochastic) machine learning with well-established methods in inductive logic programming (ILP) and rule induction to provide efficient and scalable algorithms for the classification of vast data sets. By construction, these classifications are based on the synthesis of simple rules, thus providing direct explanations of the obtained classifications. Apart from evaluating our approach on the common large scale data sets MNIST, Fashion-MNIST and IMDB, we present novel results on explainable classifications of dental bills. The latter case study stems from an industrial collaboration with Allianz Private Krankenversicherung which is an insurance company offering diverse services in Germany.

A deep learning based interval type-2 fuzzy approach for image retrieval systems

Deep learning, that one of its key benefits is automated feature extraction, has become a principal solution for computer vision. This paper presents a Deep Type-2 Beta Fuzzy (DT2F) approach for Content-Based Image Retrieval (CBIR) systems. Firstly, the suggested architecture uses InceptionResNetv2 a pre-trained deep learning model on Image-Net data as a feature extractor. Secondly, the obtained feature space is fuzzified to handle the uncertainties associated with the extracted values of deep features. Thirdly, the reduction dimensionality step is efficiently applied using a Multi-Variational Auto-Encoder (MVAE) to reduce computational complexity and achieve better performance. Ultimately, we retrieve images using the nearest neighbors rule based on type-2 fuzzy similarity to having higher proximity sensitivity. Extensive experimentations were accomplished on various image-retrieving datasets of different scales the proposed system achieved an average precision of 97.15% exceeding other state-of-the-art methods over many systems on Corel datasets, which can open the door for several hybridization breakthroughs in the area of image retrieval.

Webly Supervised Knowledge-Embedded Model for Visual Reasoning.

Visual reasoning between visual images and natural language remains a long-standing challenge in computer vision. Conventional deep supervision methods target at finding answers to the questions relying on the datasets containing only a limited amount of images with textual ground-truth descriptions. Facing learning with limited labels, it is natural to expect to constitute a larger scale dataset consisting of several million visual data annotated with texts, but this approach is extremely time-intensive and laborious. Knowledge-based works usually treat knowledge graphs (KGs) as static flattened tables for searching the answer, but fail to take advantage of the dynamic update of KGs. To overcome these deficiencies, we propose a Webly supervised knowledge-embedded model for the task of visual reasoning. On the one hand, vitalized by the overwhelming successful Webly supervised learning, we make much use readily available images from the Web with their weakly annotated texts for an effective representation. On the other hand, we design a knowledge-embedded model, including the dynamically updated interaction mechanism between semantic representation models and KGs. Experimental results on two benchmark datasets demonstrate that our proposed model significantly achieves the most outstanding performance compared with other state-of-the-art approaches for the task of visual reasoning.

Identification of Influential Nodes in Social Network: Big Data - Hadoop

Software development and associated data is the most critical factor these days. Currently, people are living in an internet world where data and related artifacts are major sets of information these days. The data is correlated with real-world data. The analysis of large datasets was done as part of the experimental analysis. The dataset for online social media like Facebook and Twitter was taken for the identification of influential nodes. The analysis of the dataset provides an overview and observation of the dataset for Facebook or Twitter. Here, in the current activity, an overview of cloud computing and big data technologies are discussed along with effective methods and approaches to resolve the problem statement. Particularly, big data technologies such as Hadoop provided by Apache for processing and analysis of Gigabyte(GB) or petabyte(PB) scale datasets are discussed for processing data in distributed and parallel data fashion. Here, the processing of large datasets is done by big data technology by implementing Apache Hadoop in online social media.

Enhanced Sub-graph Reconstruction Graph Neural Network for Recommendation

ABSTRACT Personalized recommendation can recommend items of interest to different users and is widely used in the real world. Among them, graph collaborative filtering is a method of personalized recommendation. It can enrich the connection between users and items on the basis of collaborative filtering, to learn the embedded representation of nodes more accurately. Since graph collaborative filtering is based on bipartite graphs, few exciting graph collaborative methods consider the relationships between users (or items), the message between homogeneous nodes are diluted or ignored. Predicting and constructing the relationship between users (or items) has become a challenging. To solve this problem, we propose an enhanced sub-graph reconstruction graph neural network for recommendation (SRCF), using a heterogeneous graph neural network based encoder-decoder learn potential relationships between users (or items), and reconstruct sub-graphs based on those relationships. In the proposed model, the information of user and item sub-graphs is merged with the network of graph collaborative filtering, which enhances effective information transfer between homogeneous nodes, thereby improving the model performance. We have selected a number of data sets of different scenarios and different scales to comprehensively evaluate the performance of the model, and the experimental results confirmed the superiority of our model.

Analysis-ready VCF at Biobank scale using Zarr.

Variant Call Format (VCF) is the standard file format for interchanging genetic variation data and associated quality control metrics. The usual row-wise encoding of the VCF data model (either as text or packed binary) emphasises efficient retrieval of all data for a given variant, but accessing data on a field or sample basis is inefficient. Biobank scale datasets currently available consist of hundreds of thousands of whole genomes and hundreds of terabytes of compressed VCF. Row-wise data storage is fundamentally unsuitable and a more scalable approach is needed. We present the VCF Zarr specification, an encoding of the VCF data model using Zarr which makes retrieving subsets of the data much more efficient. Zarr is a cloud-native format for storing multi-dimensional data, widely used in scientific computing. We show how this format is far more efficient than standard VCF based approaches, and competitive with specialised methods for storing genotype data in terms of compression ratios and calculation performance. We demonstrate the VCF Zarr format (and the vcf2zarr conversion utility) on a subset of the Genomics England aggV2 dataset comprising 78,195 samples and 59,880,903 variants, with a 5X reduction in storage and greater than 300X reduction in CPU usage in some representative benchmarks. Large row-encoded VCF files are a major bottleneck for current research, and storing and processing these files incurs a substantial cost. The VCF Zarr specification, building on widely-used, open-source technologies has the potential to greatly reduce these costs, and may enable a diverse ecosystem of next-generation tools for analysing genetic variation data directly from cloud-based object stores.

Improved Ant Colony Algorithm for the Split Delivery Vehicle Routing Problem

The split delivery vehicle routing problem (SDVRP) is a classic combinatorial optimization problem, which is usually solved using a heuristic algorithm. The ant colony optimization algorithm is an excellent heuristic algorithm that has been successfully applied to solve various practical problems, and it has achieved good results. However, in the existing ant colony optimization algorithms, there are issues with weak targeting of different customer selection strategies, difficulty in balancing convergence speed and global search ability, and a predisposition to become trapped in local optima. To solve these problems, this paper proposes an improved ant colony algorithm (IACA). First, in terms of customer point selection, the initial customer and noninitial customer selection strategies are proposed for different customers, and the adaptive selection threshold is designed. Second, in terms of pheromone processing, an initial pheromone distribution method based on a greedy strategy, a pheromone backtracking mechanism, and an adaptive pheromone volatile factor are proposed. Finally, based on the 2-opt local search method, vehicle path self-search and intervehicle path search are proposed to further improve the quality of the solution. This paper tests the performance of the IACA on datasets of different scales. The experimental results show that compared with the clustering algorithm, artificial bee colony algorithm, particle swarm optimization algorithm, traditional ant colony algorithm, and other algorithms, the IACA can achieve more competitive results. Specifically, compared to the path length calculated by other algorithms, the path length calculated by IACA decreased by an average of 1.58%, 4.28%, and 3.64% in small, medium, and large-scale tests, respectively.

DDI‐Transform: A neural network for predicting drug‐drug interaction events

AbstractDrug‐drug interaction (DDI) event prediction is a challenging problem, and accurate prediction of DDI events is critical to patient health and new drug development. Recently, many machine learning‐based techniques have been proposed for predicting DDI events. However, most of the existing methods do not effectively integrate the multidimensional features of drugs and provide poor mitigation of noise to get effective feature information. To address these limitations, we propose a DDI‐Transform neural network framework for DDI event prediction. In DDI‐Transform, we design a drug structure information feature extraction module and a drug bind‐protein feature extraction module to obtain multidimensional feature information. A stack of DDI‐Transform layers in the DDI‐Transform network module are then used for adaptive learning, thus adaptively selecting the effective feature information for prediction. The results show that DDI‐Transform can accurately predict DDI events and outperform the state‐of‐the‐art models. Results on different scale datasets confirm the robustness of the method.

Online continual decoding of streaming EEG signal with a balanced and informative memory buffer

Electroencephalography (EEG) based Brain Computer Interface (BCI) systems play a significant role in facilitating how individuals with neurological impairments effectively interact with their environment. In real world applications of BCI system for clinical assistance and rehabilitation training, the EEG classifier often needs to learn on sequentially arriving subjects in an online manner. As patterns of EEG signals can be significantly different for different subjects, the EEG classifier can easily erase knowledge of learnt subjects after learning on later ones as it performs decoding in online streaming scenario, namely catastrophic forgetting. In this work, we tackle this problem with a memory-based approach, which considers the following conditions: (1) subjects arrive sequentially in an online manner, with no large scale dataset available for joint training beforehand, (2) data volume from the different subjects could be imbalanced, (3) decoding difficulty of the sequential streaming signal vary, (4) continual classification for a long time is required. This online sequential EEG decoding problem is more challenging than classic cross subject EEG decoding as there is no large-scale training data from the different subjects available beforehand. The proposed model keeps a small balanced memory buffer during sequential learning, with memory data dynamically selected based on joint consideration of data volume and informativeness. Furthermore, for the more general scenarios where subject identity is unknown to the EEG decoder, aka. subject agnostic scenario, we propose a kernel based subject shift detection method that identifies underlying subject changes on the fly in a computationally efficient manner. We develop challenging benchmarks of streaming EEG data from sequentially arriving subjects with both balanced and imbalanced data volumes, and performed extensive experiments with a detailed ablation study on the proposed model. The results show the effectiveness of our proposed approach, enabling the decoder to maintain performance on all previously seen subjects over a long period of sequential decoding. The model demonstrates the potential for real-world applications.

YJMob100K: City-scale and longitudinal dataset of anonymized human mobility trajectories

Modeling and predicting human mobility trajectories in urban areas is an essential task for various applications including transportation modeling, disaster management, and urban planning. The recent availability of large-scale human movement data collected from mobile devices has enabled the development of complex human mobility prediction models. However, human mobility prediction methods are often trained and tested on different datasets, due to the lack of open-source large-scale human mobility datasets amid privacy concerns, posing a challenge towards conducting transparent performance comparisons between methods. To this end, we created an open-source, anonymized, metropolitan scale, and longitudinal (75 days) dataset of 100,000 individuals’ human mobility trajectories, using mobile phone location data provided by Yahoo Japan Corporation (currently renamed to LY Corporation), named YJMob100K. The location pings are spatially and temporally discretized, and the metropolitan area is undisclosed to protect users’ privacy. The 90-day period is composed of 75 days of business-as-usual and 15 days during an emergency, to test human mobility predictability during both normal and anomalous situations.

High-resolution physicochemical dataset of atmospheric aerosols over the Tibetan Plateau and its surroundings

Abstract. Atmospheric aerosol in the Tibetan Plateau (TP) and its surroundings has attracted significant scientific interest in recent decades due to its notable impacts on regional climatic and cryospheric changes, ecological and environmental security, and the hydrological cycle. However, our understanding of the atmospheric aerosol in this remote region is highly limited by the scarcity of available datasets owing to the extremely harsh natural conditions. This challenge has been mitigated in recent decades by establishing field observatories at typical sites within the TP and its surroundings. A continuous project initiated in 2015 aims to explore the properties and sources of atmospheric aerosols, as well as their regional differences, through multiple short-term intensive observations obtained across this vast region utilizing a suite of high-time-resolution online instruments. This paper presents a systematic and hourly scale dataset of aerosol physicochemical and optical properties at eight sites across the TP and its surroundings that is derived from the project. It includes size-resolved chemical compositions of submicron aerosols, high-resolution mass spectra and sources of organic aerosols, size distributions of particle number concentrations, particle light-scattering and light-absorption coefficients, particle light absorptions attributed to different carbonaceous substances including black carbon and brown carbon, and number concentrations of cloud condensation nuclei. In brief, atmospheric aerosols in these remote sites were all well mixed and highly aged, reflecting their dominant regional transport sources. However, the southern TP region exhibited high contributions of carbonaceous organic aerosols, neutralized bulk submicron aerosols, and a relatively high light-absorption capacity, whereas in the northern TP region, secondary inorganic species were the main contributors to the overall acidic submicron aerosols. Beyond providing insights into the regional differences in aerosol sources and properties across the TP and its surroundings, the datasets will also benefit simulations of aerosol radiative forcing and evaluations of interactions among different Earth system components in numerical models for this region. The datasets are accessible through the National Cryosphere Desert Data Center, Chinese Academy of Sciences (https://doi.org/10.12072/ncdc.NIEER.db2200.2022; Xu, 2022).