Feature Transformation Research Articles

Cross-modal similar clinical case retrieval using a modular model based on contrastive learning and k-nearest neighbor search

ObjectiveElectronic health record systems have made it possible for clinicians to use previously encountered similar cases to support clinical decision-making. However, most studies for similar case retrieval were based on single-modal data. The existing studies on cross-modal clinical case retrieval were limited. We aimed to develop a CRoss-Modal Retrieval (CRMR) model to retrieve similar clinical cases recorded in different data modalities. Materials and methodsThe publically available Medical Information Mart for Intensive Care-Chest X-ray (MIMIC-CXR) dataset was used for model development and testing. The CRMR model was designed as a modular model containing two feature extraction models, two feature transformation models, one feature transformation optimization model, and one case retrieval model. The ability to retrieve similar clinical cases recorded in different data modalities was facilitated by the use of contrastive deep learning and k-nearest neighbor search. ResultsThe average retrieval precision, denoted as AP@k, of the developed CRMR model, were 76.9 %@5, 76.7 %@10, 76.5 %@20, 76.3 %@50, and 77.9 %@100, respectively. Here k is the number of similar cases returned after retrieval. The average retrieval time varied from 0.013 ms to 0.016 ms with k varying from 5 to 100. Moreover, the model can retrieve similar cases with the same multiple radiographic manifestations as the query case. DiscussionThe CRMR model has shown promising cross-modal retrieval performance in clinical case analysis, with the potential for future scalability and improvement in handling diverse disease types and data modalities. The CRMR model has promising potential to aid clinicians in making optimal and explainable clinical decisions.

Feature Bagging with Nested Rotations (FBNR) for anomaly detection in multivariate time series

Detecting anomalies in multivariate time series poses a significant challenge across various domains. The infrequent occurrence of anomalies in real-world data, as well as the lack of a large number of annotated samples, makes it a complex task for classification algorithms. Deep Neural Network approaches, based on Long Short-Term Memory (LSTMs), Autoencoders, and Variational Autoencoders (VAEs), among others, prove effective with handling imbalanced data. However, the same does not follow when such algorithms are applied on multivariate time-series, as their performance degrades significantly. Our main hypothesis is that the above is due to anomalies stemming from a small subset of the feature set. To mitigate the above issues in the multivariate setting, we propose forming an ensemble of base models by combining different feature selection and transformation techniques. The proposed processing pipeline includes applying a Feature Bagging techniques on multiple individual models, which considers separate feature subsets for each specific model. These subsets are then partitioned and transformed using multiple nested rotations derived from Principal Component Analysis (PCA). This approach aims to identify anomalies that arise from only a small portion of the feature set while also introduces diversity by transforming the subspaces. Each model provides an anomaly score, which are then aggregated, via an unsupervised decision fusion model. A semi-supervised fusion model was also explored, in which a Logistic Regressor was applied on the individual model outputs. The proposed methodology is evaluated on the Skoltech Anomaly Benchmark (SKAB), containing multivariate time series related to water flow in a closed circuit, as well as the Server Machine Dataset (SMD), which was collected from a large Internet company. The experimental results reveal that the proposed ensemble technique surpasses state-of-the-art algorithms. The unsupervised approach demonstrated a performance improvement of 2% for SKAB and 3% for SMD, compared to the baseline models. In the semi-supervised approach, the proposed method achieved a minimum of 10% improvement in terms of anomaly detection accuracy.

Live feature transformation scale extraction of face images based on deep learning algorithm

In view of the low extraction accuracy and efficiency of the current feature extraction methods of face images, a method for extracting the scale of live feature transformation of face images based on deep learning algorithm is proposed. First, the deep learning method is used to complete the denoising of face images; then the Gabor wave filter is used to decompose the face signal, and it is input into the deep subspace model to complete the extraction of feature transformation scale; finally, the live feature transformation scale extraction of face images is completed based on (particle PSO swarm optimization algorithm). Experimental results show that the proposed method for extracting face image features has higher accuracy, faster recognition speed and better overall application effect.

Exploring Feature Relationships in Brain Stroke Data Using Polynomial Feature Transformation and Linear Regression Modeling

A Cerebral vascular accident, commonly known as a stroke, is a pathological condition that impacts the brain due to the rupture of capillaries. It occurs when there is a disturbance in the typical blood circulation and essential physiological processes of the brain. Stroke prediction plays a crucial role in early diagnosis and intervention, potentially improving patient outcomes. This paper proposes a machine learning model that leverages polynomial feature transformation and linear regression modeling for stroke prediction. The model addresses the challenge of capturing non-linear relationships between features and the target variable while maintaining interpretability. The proposed approach involves preprocessing data by separating categorical and numerical features, applying one-hot encoding to categorical features, and generating polynomial features up to the second degree for numerical features. This tailored preprocessing is facilitated by a Column Transformer. For model development, a machine learning pipeline is constructed, splitting the data into training and testing sets. Despite utilizing polynomial features, linear regression is employed as the final model, allowing for the capture of both linear and non-linear relationships while maintaining interpretability. This work contributes to stroke prediction by offering a balanced approach that considers model complexity and interpretability, showcasing the potential of linear regression with polynomial features for accurate predictions and insights into feature-target relationships. The proposed model exhibited superior performance compared to other existing models, achieving a remarkable testing accuracy of 99.2%.

Optimizing a Dynamic Vehicle Routing Problem with Deep Reinforcement Learning: Analyzing State-Space Components

Background: The dynamic vehicle routing problem (DVRP) is a complex optimization problem that is crucial for applications such as last-mile delivery. Our goal is to develop an application that can make real-time decisions to maximize total performance while adapting to the dynamic nature of incoming orders. We formulate the DVRP as a vehicle routing problem where new customer requests arrive dynamically, requiring immediate acceptance or rejection decisions. Methods: This study leverages reinforcement learning (RL), a machine learning paradigm that operates via feedback-driven decisions, to tackle the DVRP. We present a detailed RL formulation and systematically investigate the impacts of various state-space components on algorithm performance. Our approach involves incrementally modifying the state space, including analyzing the impacts of individual components, applying data transformation methods, and incorporating derived features. Results: Our findings demonstrate that a carefully designed state space in the formulation of the DVRP significantly improves RL performance. Notably, incorporating derived features and selectively applying feature transformation enhanced the model’s decision-making capabilities. The combination of all enhancements led to a statistically significant improvement in the results compared with the basic state formulation. Conclusions: This research provides insights into RL modeling for DVRPs, highlighting the importance of state-space design. The proposed approach offers a flexible framework that is applicable to various variants of the DVRP, with potential for validation using real-world data.

On filling the intra-class and inter-class gaps for few-shot segmentation

Few-shot segmentation (FSS) has made remarkable success through prototypical learning. However, owing to the scarcity of support data, FSS methods continue to suffer from huge intra-class and inter-class gaps. In this paper, we introduce a unified network, termed FGNet++, aimed at addressing both gaps. FGNet++ comprises a Self-Adaptive Module (SAM) to emphasize the query features and generate enhanced prototypes for self-alignment. These prototypes contain the intrinsic information of each query sample, thereby mitigating the intra-class appearance gaps. Moreover, we augment SAM with a Feature Transformation Module (FTM) to further reduce intra-class appearance discrepancies. On the other hand, we introduce an Inter-class Feature Separation Module (IFSM) dedicated to bridge the inter-class gap. With a B-SLIC based background loss and a cross-category loss for training, IFSM makes it easy to distinguish the feature space of the target class from other non-target classes. Furthermore, to assess the generality of our approach, we extend our work to 3D point cloud few-shot segmentation and present FGNet-3D. Experimental results demonstrate that our method successfully mitigates both intra-class and inter-class gaps in FSS through SAM and IFSM, respectively, and achieves state-of-the-art performances on multiple datasets compared with previous approaches.

Sound source localization via distance metric learning with regularization

Sound source localization (SSL) or simply direction of arrival (DOA) classification is an important ingredient in acoustic applications. Traditional model-based algorithms are susceptible to the effects of noise and reverberation, while data-driven deep learning algorithms maintain strong performance across a variety of acoustic circumstances, but typically require a large amount of labeled data. Nevertheless, the existing datasets for SSL are not sufficiently big and diverse to achieve the full potential of deep learning algorithms. Then, it is an imperative work to develop a non-data-hungry algorithm of SSL using small or medium data volume. To this end, we propose a regularized distance metric learning algorithm, that is, by means of the kernel method, we design a nonlinear feature transformation from two aspects: feature points and feature distributions. It transforms the data into a new feature space that brings features of the same class as close as possible and removes features of different classes as far away as possible, which can significantly improve the output of a DOA classifier that follows. Experimental results show that the proposed algorithm outperforms deep learning algorithms in diverse acoustic conditions.

The essence and assessment of modern industrial policy

The state of modern economy is exposed to significant challenges from geopolitical risks and transformation of the process of industrial development, which requires determining the conditions for implementing modern industrial policy and changing directions and priorities of industrial development. Improvement of the ways of production, development of the system of social institutions takes place when solving the issues related to the development of science underlying the formation of new industries. The subject of the study is the transformation of industrial policy features in modern economy at various levels of interaction between government and business. The article examines the changes in industrial development and corresponding transformation of industrial policy tools in import substitution development. The study aims at determining the relationships between the directions and levels of implementing industrial policy in accordance with system analysis and identifying the key features in the relationship between business structures and the state. The theories of industrial development and the concepts of resource and process approaches to industrial development and implementation of innovative industrial policy provide the methodological basis of the study. The study uses the methods of system analysis and synthesis in identifying the key relationships between business structures and the state in achieving the goals of industrial development. The study defines the transformation of types, tools and system of indicators of industrial policy under the sanction pressure on the economy, identifies the possibilities of using the matrix system in managing industrial development, implemented at three levels (macro-, meso- and microeconomic) in interaction of business structures and public authorities, assesses the possibility of forming and conducting conservative industrial policy in the lack of scientific research funding. The findings can be used in assessing the priorities, directions and transformation of the tools used to implement an innovative model of industrial policy in the Russian economy, in the development of import substitution and formation of new industries, to form an effective interaction of innovation process and the process of industrial development of the national economy.

Pseudo Oversampling Based on Feature Transformation and Fuzzy Membership Functions for Imbalanced and Overlapping Data

Class imbalance in data poses challenges for classifier learning, drawing increased attention in data mining and machine learning. The occurrence of class overlap in real-world data exacerbates the learning difficulty. In this paper, a novel pseudo oversampling method (POM) is proposed to learn imbalanced and overlapping data. It is motivated by the point that overlapping samples from different classes share the same distribution space, and therefore information underlying in majority (negative) overlapping samples can be extracted and used to generate additional positive samples. A fuzzy logic-based membership function is defined to assess negative overlaps using both local and global information. Subsequently, the identified negative overlapping samples are shifted into the positive sample region by a transformation matrix, centered around the positive samples. POM outperforms 15 methods across 14 datasets, displaying superior performance in terms of metrics of <i>G<sub>m</sub></i>, <I>F</I><sub>1</sub> and <I>AUC</I>.

Hybrid convolutional neural networks in human resource recommendation

The study proposes a new algorithm combining a gradient boosting tree algorithm with a hybrid convolutional neural network in order to design a better human resource recommendation algorithm to solve the problem of employment difficulties and recruitment difficulties. The algorithm combines the excellent feature transformation ability of gradient boosting trees with the excellent classification ability of hybrid convolutional neural networks, complementing the shortcomings of each of the two algorithms. The outcomes showed that the algorithm performed best with the learning rate set to 0.3 and the maximum tree depth set to 3. The algorithm now has the lowest loss percentage and highest F1-Score value. The maximum-median hybrid pooling approach used in the study had considerable improvements over the algorithm’s pooling strategy (PS), and it had the greatest recall and F1-Score values of all pooling strategies (0.8108 and 0.7418, respectively). The new algorithm outperformed the gradient boosting tree algorithm and the hybrid convolutional neural network algorithm in terms of recall and F1-Score values, and regardless of the length of the job recommendation, the recall and F1-Score values of the new algorithm were consistently higher than those of the old algorithm. The recall and F1-Score values of the new algorithm, with a job recommendation length of 70, were 0.8198 and 0.7432, respectively, both greater than those of the other algorithms, according to a comparison of it with other conventional HR recommendation algorithms. The study’s newly created algorithm increases the efficacy and precision of HR suggestions.

Developing and Implementing an Artificial Intelligence (AI)-Driven System For Electricity Theft Detection

Electricity theft is a significant challenge for utility companies worldwide, leading to substantial economic losses and inefficiencies in power distribution. Traditional methods of detecting electricity theft, such as manual inspections and routine audits, are often inefficient and ineffective. To address this issue, this study aims to develop and implement an artificial intelligence (AI)-driven system for electricity theft detection. Methodology used are data collection, data analysis, feature selection with Chi-Square, feature transformation with Principal Component Analysis (PCA), Support Vector Machine (SVM) and model for electricity theft detection. To achieve this, a Particle Swarm Optimization Algorithm (PSO) was applied to improve training performance of the SVM, using data of meter recharge information collected from Enugu Electricity Distribution Company (EEDC). The system effectiveness is validated through extensive testing using real-world data from various regions and scenarios, demonstrating its robustness and adaptability. The system result considering FDR reported that 0.11 was achieved for the particle swarm based SVM model. When TPR was considered for analysis, it was observed that particle swarm based SVM attained a score of 0.89. In addition, Particle swarm based SVM attained PPV of 0.895. In terms of accuracy, the particle swarm based SVM reported an accuracy of 0.857. The result showed that the particle swarm based SVM performed better from the system validation achieved through comparative analysis, hence it is recommended for use to develop the new software for energy theft investigation. The implementation of this AI-driven solution offers numerous benefits, including enhanced detection accuracy, reduced operational costs, and improved overall efficiency of power distribution networks. Moreover, it enables utility companies to take proactive measures to prevent theft, ensuring a more reliable and secure electricity supply for consumers.

A district-level building electricity use profile simulation model based on probability distribution inferences

District-level building energy systems play a significant role in urban energy networks in the future. Understanding the key distributive features of district electricity use profiles is essential for the optimal planning and design of energy networks. Due to the diversity of building electricity use characteristics, the district-level electricity use profile exhibits a prominent “peak staggering effect.” Current physics-based and statistical models cannot fully represent realistic distributions and the uncertainties of district profiles. Thus, it is critical to quantitatively investigate the changing patterns and distributive features of electricity use profiles at various district levels. This paper proposes a novel approach for district building electricity use profile simulation. Probability distribution inference methods integrating Gaussian Mixture Model (GMM)/lognorm distribution fitting, singular value decomposition (SVD)-based feature transformation, and distribution addition theorems have been proposed to generate the feature parameters of electricity use profiles at various district scales, thus generating simulated district electricity use profiles. The performance of the proposed model was validated using engineering-informed metrics, including peak demands, load duration curves, and standard deviations of the load parameters. The results of the case study suggest that the average relative error of the 99 % peak demand is reduced from 17.60 % in the baseline model to 3.48 % in the proposed model, the average relative error of the duration of 2Qm reduced from 40.82 % in the baseline model to 0.99 % in the proposed model, and the average relative error of the standard deviation of load parameters was reduced from >100 % in the baseline model to <35 % in the proposed model. The results indicate a better quantification of district electricity use distributions and uncertainties, providing practical tools to support the capacity design and optimization of integrated district energy systems.

AmazingFT: A Transformer and GAN-Based Framework for Realistic Face Swapping

Current face-swapping methods often suffer from issues of detail blurriness and artifacts in generating high-quality images due to the inherent complexity in detail processing and feature mapping. To overcome these challenges, this paper introduces the Amazing Face Transformer (AmazingFT), an advanced face-swapping model built upon Generative Adversarial Networks (GANs) and Transformers. The model is composed of three key modules: the Face Parsing Module, which segments facial regions and generates semantic masks; the Amazing Face Feature Transformation Module (ATM), which leverages Transformers to extract and transform features from both source and target faces; and the Amazing Face Generation Module (AGM), which utilizes GANs to produce high-quality swapped face images. Experimental results demonstrate that AmazingFT outperforms existing state-of-the-art (SOTA) methods, significantly enhancing detail fidelity and occlusion handling, ultimately achieving movie-grade face-swapping results.

A Lightweight Insulator Defect Detection Model Based on Drone Images

With the continuous development and construction of new power systems, using drones to inspect the condition of transmission line insulators has become an inevitable trend. To facilitate the deployment of drone hardware equipment, this paper proposes IDD-YOLO (Insulator Defect Detection-YOLO), a lightweight insulator defect detection model. Initially, the backbone network of IDD-YOLO employs GhostNet for feature extraction. However, due to the limited feature extraction capability of GhostNet, we designed a lightweight attention mechanism called LCSA (Lightweight Channel-Spatial Attention), which is combined with GhostNet to capture features more comprehensively. Secondly, the neck network of IDD-YOLO utilizes PANet for feature transformation and introduces GSConv and C3Ghost convolution modules to reduce redundant parameters and lighten the network. The head network employs the YOLO detection head, incorporating the EIOU loss function and Mish activation function to optimize the speed and accuracy of insulator defect detection. Finally, the model is optimized using TensorRT and deployed on the NVIDIA Jetson TX2 NX mobile platform to test the actual inference speed of the model. The experimental results demonstrate that the model exhibits outstanding performance on both the proprietary ID-2024 insulator defect dataset and the public SFID insulator dataset. After optimization with TensorRT, the actual inference speed of the IDD-YOLO model reached 20.83 frames per second (FPS), meeting the demands for accurate and real-time inspection of insulator defects by drones.

DeepQCD: An end-to-end deep learning approach to quickest change detection

This paper aims to generalize the quickest change detection (QCD) framework via a data-driven approach. To this end, a generic neural network architecture is proposed for the QCD task, composed of feature transformation, recurrent, and dense layers. The neural network is trained end-to-end to learn the change detection rule directly from data without needing the knowledge of probabilistic data models. Specifically, the feature transformation layers can perform a broad range of operations including feature extraction, scaling, and normalization. The recurrent layers keep an internal state summarizing the time-series data seen so far and update the state as new information comes in. Finally, the dense layers map the internal state into a decision statistic, defined as the posterior probability that a change has taken place. Comparisons with the existing model-based QCD algorithms demonstrate the power of the proposed data-driven approach, called DeepQCD, under several scenarios including transient changes and temporally correlated data streams. Experiments with real-world data illustrate superior performance of DeepQCD compared to state-of-the-art algorithms in real-time anomaly detection over surveillance videos and real-time attack detection over Internet of Things (IoT) networks.

Learning a multi-cluster memory prototype for unsupervised video anomaly detection

In recent years, there has been rapid development in video anomaly detection (VAD). The previous methods ignored the differences between normal videos and only emphasized learning the commonalities of normal videos. In order to improve the performance of anomaly detection, we delve into the spatial distribution of normal video features and utilize their differences for clustering, leading to more minor reconstruction errors for normal videos and more significant reconstruction errors for abnormal videos. To achieve this goal, we introduce a Multi-Cluster Memory Prototype framework (MCMP) for VAD, which explores the coarse-grained and fine-grained information from video snippet features simultaneously to learn a memory prototype, thereby significantly improving the ability to discriminate abnormal events in complex scenes. First, a video feature clustering method that employs contrastive learning is introduced to group samples sharing similar fine-grained features. Second, the memory mechanism is used to capture the feature distribution of normal samples. Lastly, the Gaussian filter feature transformation method is introduced to make normal and abnormal features more distinguishable. The frame level AUC of MCMP on ShanghaiTech and UCF-Crime benchmark datasets has increased by 1.26% and 0.45% compared to state-of-the-art methods. Our code is publicly available athttps://github.com/WuIkun5658/MCMP.

Micro-displacement measurement of optical interference with two-ring conjugated vortices

In this paper, we propose a two-ring conjugated vortex beam interferometer. The inner ring and outer ring of the vortex beams has different signs and topological charges. When combined, they produce an intricate pattern of interference fringes that resemble a flower petal, featuring both concentric inner and outer rings which rotates with different directions. We use the scale-invariant feature transformation algorithm to find the feature points. Then we use these points to measure the rotation angle of the outer ring relative to the inner ring and calculate the displacement based on it. We realize the measurement of sub-nanometer displacement. The displacement resolution is 0.38° nm−1. It is improved by about 12.55% compared with only the inner ring. The displacement corresponding to the minimum detectable angle of CCD is reduced from 0.080 nm to 0.060 nm. It will be a promising method in precision machining micro-vibration sensing, surface unevenness sensing, and microbial movement sensing and high-precision measurement of micro-displacement.

Research on Vibration Control Regarding Mechanical Coupling for Maglev Trains with Experimental Verification

The electromagnet module, as a fundamental component providing levitation force for maglev trains, plays a crucial role in ensuring the stability of train operation. However, vibrations can easily occur due to the mechanical coupling between the two suspension points of the electromagnet module. To reveal the inherent instability of the system and the coupling relationship between the state variables, a state-space equation that considers the mechanical coupling between the two suspension points is established. Furthermore, a differential control algorithm based on geometric feature transformation is proposed to mitigate the structural coupling vibration. Simulation experiments are conducted to compare the dynamic characteristics of the system before and after implementing the improvement algorithm under complex conditions. At the same time, the influence of control parameters on electromagnetic vibration was analyzed, focusing particularly on vibrations resulting from parameter mismatch, offering crucial insights for enhancing system stability. Additionally, suspension tests are carried out on the high-speed double bogie test platform in the Key Laboratory of Hunan Province to further validate the effectiveness of the proposed algorithm. The proposed control framework is both effective and concise, making it easy to implement in engineering applications. This research holds significant practical value in improving the stability of maglev trains.

DNIM: Deep-sea netting intelligent enhancement and exposure monitoring using bio-vision

Intelligent monitoring of deep-sea nets is affected by light attenuation, light scattering, and limited dynamic range factors of the camera, which can cause color shift, low visibility, and over/underexposure in monitoring, resulting in reduced farming efficiency, and misdetection of biological behavior. Therefore, we propose a method for underwater net tank scene enhancement and exposure using multicolor space and bio-vision, called DNIM. Specifically, we first propose an intelligent enhancement and exposure monitoring method for deep-sea nets using bio-vision, which eliminates the effect of exposure on underwater imaging by introducing a bio-vision system that uses a brightness transformation function to generate a series of exposure images. Inspired by the image fusion method, we designed a color correction and visibility recovery method using bio-vision. In the Lab color model, color shifts in bio-vision are corrected by guiding feature transformations with contrast-constrained adaptive histogram equalization; in the RGB color model, depth information is recovered by creating a linear model for the depth of field of underwater blurred images based on the characteristics of the bio-vision system. To obtain more qualified results, we also propose an efficient perceptual fusion module to mix the output of visibility recovery and color correction. Finally, we propose two deep-sea net tank culture datasets: salmon bait-eating behavior dataset (SBBD) and salmon non-bait-eating behavior dataset (SNBD). By comparing the experimental analysis in four real scenario datasets, our DNIM achieved superior results compared to 16 advanced underwater enhancement methods. The code is publicly available at: https://github.com/An-Shunmin/DNIM.

A study of online academic risk prediction based on neural network multivariate time series features

SummaryNeural networks are becoming increasingly widely used in various fields, especially for academic risk forecasts. Academic risk prediction is a hot topic in the field of big data in education that aims to identify and help students who experience great academic difficulties. In recent years, the use of machine learning algorithms and deep learning algorithms to achieve academic risk prediction has garnered increased attention and development. However, most of these studies use nontime series data as features for prediction, which are slightly insufficient in terms of timeliness. Therefore, this article focuses on time series data features that are more expressive of changes in students' learning status and uses multivariate time series data as predictive features. This article proposes a method based on multivariate time series features and a neural network to predict academic risk. The method includes three steps: first, the multivariate time series feature is extracted from the interaction records of the students' online learning platforms; second, the multivariate time series feature transformation model ROCKET is applied to convert the multivariate time series feature into a new feature; third, the new feature is converted into a final prediction result using a linear classification model. Comparative tests show that the proposed method has high effectiveness.