High-quality Training Samples Research Articles

The Research of Intra-Pulse Modulated Signal Recognition of Radar Emitter under Few-Shot Learning Condition Based on Multimodal Fusion

Radar radiation source recognition is critical for the reliable operation of radar communication systems. However, in increasingly complex electromagnetic environments, traditional identification methods face significant limitations. These methods often struggle with high noise levels and diverse modulation types, making it difficult to maintain accuracy, especially when the Signal-to-Noise Ratio (SNR) is low or the available training data are limited. These difficulties are further intensified by the necessity to generalize in environments characterized by a substantial quantity of noisy, low-quality signal samples while being constrained by a limited number of desirable high-quality training samples. To more effectively address these issues, this paper proposes a novel approach utilizing Model-Agnostic Meta-Learning (MAML) to enhance model adaptability in few-shot learning scenarios, allowing the model to quickly learn with limited data and optimize parameters effectively. Furthermore, a multimodal fusion neural network, DCFANet, is designed, incorporating residual blocks, squeeze and excitation blocks, and a multi-scale CNN, to fuse I/Q waveform data and time–frequency image data for more comprehensive feature extraction. Our model enables more robust signal recognition, even when the signal quality is severely degraded by noise or when only a few examples of a signal type are available. Testing on 13 intra-pulse modulated signals in an Additive White Gaussian Noise (AWGN) environment across SNRs ranging from −20 to 10 dB demonstrated the approach’s effectiveness. Particularly, under a 5−way5−shot setting, the model achieves high classification accuracy even at −10dB SNR. Our research underscores the model’s ability to address the key challenges of radar emitter signal recognition in low-SNR and data-scarce conditions, demonstrating its strong adaptability and effectiveness in complex, real-world electromagnetic environments.

Landsat image classification using a deep learning model and multiple-source training samples

ABSTRACT Public release of the entire Landsat data archive and cloud-based geocomputation platforms have greatly facilitated land-cover mapping worldwide. The acquisition of training samples presents a significant challenge in mapping. In this study, we have developed an approach based on transfer learning to creatively use existing global land-cover products and training samples to resolve the training sample issue. We pretrained a deep neural network (DNN) model using low-quality samples automatically generated from existing global land-cover products. The pretrained model was then fine-tuned using high-quality training samples gathered from the training samples shared by authors of existing studies. We implemented this approach to generate a land-cover map for Beijing City, China, in 2015. Using the fine-tuned DNN model, we achieved an overall accuracy (OA) of 86.4% and a kappa coefficient of 0.796, based on independent validation samples. The accuracy was the same and even surpassed that of existing land-cover maps for Beijing. The fine-tuned DNN outperformed the random forest (RF) (OA = 70.2%, kappa = 0.588) and support vector machine (SVM) models (OA = 68.7%, kappa = 0.555) using the same training samples, showing that the need for training samples in land-cover classification can be met by combining deep learning models, existing global land-cover products, and recycled training samples.

Novel physics-informed optimization framework for complex multi-physics problems: Implementation for a sweeping gas membrane distillation module

The application of advanced machine learning algorithms such as deep learning is limited for the surrogate-based process optimization of complex multi-physics problems because high-quality experimental and numerical training samples required for deep-learning model training are expensive and scarce. To address this issue, in this study, a novel physics-informed process optimization (PIPO) framework was introduced. In the first step, surrogate models based on conventional neural networks (NN) were trained using a few available high-quality training samples. In the second step, the trained NN models were coupled to the process optimization solver in which the loss of physical laws was added to the optimizer’s objective function to find optimal design points that satisfy the laws of physics. As a result, the generalization performance of the framework was greatly improved for design targets outside the training range of NN models. PIPO is substantially different from the physics-informed neural networks where the loss of physics is added to the loss function used during NN model training. The PIPO framework was used to optimize a sweeping gas membrane distillation (SGMD) module. Eight input design variables, including process and geometrical parameters, were optimized for different challenging targets to achieve the best SGMD performance in terms of ammonia recovery ratio and concentration. It was shown that for noticeably few training samples of 68 experiments, the proposed framework was able to achieve the optimization targets within a reasonable computational cost. The optimum designs were verified and analyzed in detail by high-resolution computational fluid dynamics models.

A spatial scene reconstruction framework in emergency response scenario

Rapid and accurate acquisition and analysis of information is crucial for emergency management, but traditional methods have limitations such as incomplete information acquisition and slow processing speed. The natural language oriented spatial scene reconstruction method provides a new solution for emergency management, but existing generative models have limited understanding of spatial relationships and lack high-quality training samples. To address these issues, this paper proposes a novel spatial scene reconstruction framework. Specifically, the BERT based spatial information knowledge graph extraction method is used to encode the input text, label and classify the encoded text, identify spatial objects and relationships in the text, and accurately extract spatial information. Additionally, a large number of manual experiments were conducted to explore quantitative biases in human spatial cognition, and based on the obtained biases, a greedy resolution method based on cost functions was used to fine tune the layout of conflicting spatial objects and solve the conflicting spatial information in the spatial information knowledge graph. Finally, use graph convolutional neural networks to obtain scene knowledge graph embeddings that consider spatial constraints. In addition, a high-quality training sample set of “text-scene-knowledge graph” was constructed.

Integration of Sentinel-1 and Sentinel-2 Data for Ground Truth Sample Migration for Multi-Temporal Land Cover Mapping

Reliable and up-to-date training reference samples are imperative for land cover (LC) classification. However, such training datasets are not always available in practice. The sample migration method has shown remarkable success in addressing this challenge in recent years. This work investigated the application of Sentinel-1 (S1) and Sentinel-2 (S2) data in training sample migration. In addition, the impact of various spectral bands and polarizations on the accuracy of the migrated training samples was also assessed. Subsequently, combined S1 and S2 images were classified using the Support Vector Machines (SVM) and Random Forest (RF) classifiers to produce annual LC maps from 2017 to 2021. The results showed a higher accuracy (98.25%) in training sample migrations using both images in comparison to using S1 (87.68%) and S2 (96.82%) data independently. Among the LC classes, the highest accuracy in migrated training samples was found for water, built-up, bare land, grassland, cropland, and wetland. Inquiries on the efficiency of different spectral bands and polarization used in training sample migration showed that bands 4 and 8 and VV polarization in the water class were more important, while for the wetland class, bands 5, 6, 7, 8, and 8A together with VV polarization showed superior performance. The results showed that the RF classifier provided better performance than the SVM (higher overall, producer, and user accuracy). Overall, our findings suggested that shared use of S1 and S2 data can be used as a suitable means for producing up-to-date and high-quality training samples.

Unsupervised dynamic prognostics for abnormal degradation of lithium-ion battery

Lithium-ion batteries may suffer an abnormal degradation defined by a significantly accelerated performance drop after a period of linear and low-rate degradation, resulting in severe danger to operational safety and reliability. Existing supervised data-driven prognostics for abnormal degradation rely heavily on adequate high-quality training samples, thus hindering their real-world utilization. Therefore, this paper develops an unsupervised dynamic prognostics framework that is dynamically executed as the cycle number increases to provide accurate and timely warnings for abnormal degradation. In the framework, an extended asymmetric quantum clustering (EAQC) can preliminarily identify the risky battery with the abnormal degradation trend. Its use of multi-dimensional feature degradation rates and potential function for clustering makes it outperform other density-based clustering methods. Then, for the identified battery, a time-varying double-layer autoregression (TVDLAR) can accurately predict its knee point of abnormal degradation. TVDLAR produces much earlier warning than other autoregression methods, which owes to its modeling ability of the time-varying correlation in battery historical degradation data. Through applying to two experimental datasets consisting of 174 lithium-ion batteries of different types under various working conditions, the framework is proven highly effective and shows significant superiority over some alternative approaches; all abnormal degradation can be timely warned before the corresponding knee points, and the average earlier warning cycles are 86 and 116, respectively. Compared with supervised data-driven methods, the proposed unsupervised dynamic prognostics framework has the advantages of low data requirement, low computational consumption, and good interpretability, indicating its substantial potential for application.

Stellar Loci. VII. Photometric Metallicities of 5 Million FGK Stars Based on GALEX GR6+7 AIS and Gaia EDR3

We combine photometric data from GALEX GR6+7 All-Sky Imaging Survey and Gaia Early Data Release 3 with stellar parameters from the SAGA and PASTEL catalogs to construct high-quality training samples for dwarfs (0.4 < BP − RP < 1.6) and giants (0.6 < BP − RP < 1.6). We apply careful reddening corrections using empirical temperature- and extinction-dependent extinction coefficients. Using the two samples, we establish a relationship between stellar loci (near-ultraviolet (NUV)−BP versus BP − RP colors), metallicity, and M G . For a given BP − RP color, a 1 dex change in [Fe/H] corresponds to an approximately 1 magnitude change in NUV − BP color for solar-type stars. These relationships are employed to estimate metallicities based on NUV − BP, BP − RP, and M G . Thanks to the strong metallicity dependence in the GALEX NUV band, our models enable a typical photometric-metallicity precision of approximately σ [Fe/H] = 0.11 dex for dwarfs and σ [Fe/H] = 0.17 dex for giants, with an effective metallicity range extending down to [Fe/H] = −3.0 for dwarfs and [Fe/H] = −4.0 for giants. We also find that the NUV-band-based photometric-metallicity estimate is not as strongly affected by carbon enhancement as previous photometric techniques. With the GALEX and Gaia data, we have estimated metallicities for about 5 million stars across almost the entire sky, including approximately 4.5 million dwarfs and 0.5 million giants. This work demonstrates the potential of the NUV band for estimating photometric metallicities, and sets the groundwork for utilizing the NUV data from space telescopes such as the upcoming Chinese Space Station Telescope.

Few-shot concealed object detection in sub-THz security images using improved pseudo-annotations

In this research, we explore the few-shot object detection application for identifying concealed objects in sub-terahertz security images, using fine-tuning based frameworks. To adapt these machine learning frameworks for the (sub-)terahertz domain, we propose an innovative pseudo-annotation method to augment the object detector by sourcing high-quality training samples from unlabeled images. This approach employs multiple one-class detectors coupled with a fine-grained classifier, trained on supporting thermal-infrared images, to prevent overfitting. Consequently, our approach enhances the model’s ability to detect challenging objects (e.g., 3D-printed guns and ceramic knives) when few-shot training examples are available, especially in the real-world scenario where images of concealed dangerous items are scarce.

A 3D BUILDING INDOOR-OUTDOOR BENCHMARK FOR SEMANTIC SEGMENTATION

Abstract. Both machine learning (ML) and deep learning (DL) algorithms require high-quality training samples as well as precise and thorough annotations in order to work effectively. The 3D building indoor-outdoor dataset (BIO dataset), which is a highly accurate, high level of detail, and high coverage dataset for 3D building point cloud and mesh semantic segmentation, is established as a canonical benchmark dataset. It contains 100 building models, in which building structural elements are annotated into 11 semantic categories. Each building in this dataset has an average of 75,587 triangular faces, and the total area of the dataset is 481,769 square meters. Furthermore, semantic segmentation of the dataset was carried out using the Random Forest ML algorithm to verify the dataset’s accessibility. A weighted F1 score of 96.64% was obtained with 10% of the segments of each building randomly chosen as training data. For applications involving building geometry data, the BIO dataset can support a broad class of recently developed ML and DL methodologies.

BlockMatch: A Fine-Grained Binary Code Similarity Detection Approach Using Contrastive Learning for Basic Block Matching

Binary code similarity detection (BCSD) plays a vital role in computer security and software engineering. Traditional BCSD methods heavily rely on specific features and necessitate rich expert knowledge, which are sensitive to code alterations. To improve the robustness against minor code alterations, recent research has shifted towards machine learning-based approaches. However, existing BCSD approaches mainly focus on function-level matching and face challenges related to large batch optimization and high quality sample selection at the basic block level. To overcome these challenges, we propose BlockMatch, a novel fine-grained BCSD approach that leverages natural language processing (NLP) techniques and contrastive learning for basic block matching. We treat instructions of basic blocks as a language and utilize a DeBERTa model to capture relative position relations and contextual semantics for encoding instruction sequences. For various operands in binary code, we propose a root operand model pre-training task to mitigate semantic missing of unseen operands. We then employ a mean pooling layer to generate basic block embeddings for detecting binary code similarity. Additionally, we propose a contrastive training framework, including a block augmentation model to generate high-quality training samples, improving the effectiveness of model training. Inspired by contrastive learning, we adopt the NT-Xent loss as our objective function, which allows larger sample sizes for model training and mitigates the convergence issues caused by limited local positive/negative samples. By conducting extensive experiments, we evaluate BlockMatch and compare it against state-of-the-art approaches such as PalmTree and SAFE. The results demonstrate that BlockMatch achieves a recall@1 of 0.912 at the basic block level under the cross-compiler scenario (pool size = 10), which outperforms PalmTree (0.810) and SAFE (0.798). Furthermore, our ablation study shows that the proposed contrastive training framework and root operand model pre-training task help our model achieve superior performance.

A 3D INDOOR-OUTDOOR BENCHMARK DATASET FOR LoD3 BUILDING POINT CLOUD SEMANTIC SEGMENTATION

Abstract. Deep learning (DL) algorithms require high quality training samples as well as accurate and thorough annotations to work effectively. Up until now a limited number of datasets are available to train DL techniques for semantic segmentation of 3D building point clouds, except a few ones focusing on specific categories of constructions (e.g., cultural heritage buildings). This paper presents a new 3D Indoor/Outdoor building dataset (BIO dataset), which is aimed to provide a highly accurate, detailed, and comprehensive dataset to be used for applications related to sematic classification of buildings based on point clouds and meshes. This benchmark dataset contains 100 building models generated from existing polygonal models and belonging to different categories. These include commercial buildings, residential houses, industrial and institutional buildings. Structural elements of buildings are annotated into 11 semantic categories, following standards from IFC and CityGML. To verify the applicability of the BIO dataset for the semantic segmentation task, it has been successfully tested by using one machine learning technique and four different DL algorithms.

Classification and detection of diabetic retinopathy based on multi-scale shallow neural network

&lt;p class="15"&gt;The high-quality annotated training samples in medical image processing have limited the development of deep neural networks in their field. This paper designs and proposes an integrated method for classifying and detecting diabetic retinopathy based on a multi-scale shallow neural network. The method consists of multiple shallow neural network base learners, which extract pathological features under different receptive fields. The integrated learning strategy proposed is used to optimize the integration and finally realize the classification and detection of diabetic retinopathy. In addition, to verify the effectiveness of the method in this paper on a small sample data-set, based on the two-dimensional entropy of the image, multiple sub-datasets are constructed for verification. The results show that, compared with the existing methods, the integrated method for the classification and detection of diabetic retinopathy proposed in this paper has a good detection effect on a small sample data-set.&lt;/p&gt;

Conditional synthetic food image generation

Generative Adversarial Networks (GAN) have been widely investigated for image synthesis based on their powerful representation learning ability. In this work, we explore the StyleGAN and its application of synthetic food image generation. Despite the impressive performance of GAN for natural image generation, food images suffer from high intra-class diversity and inter-class similarity, resulting in overfitting and visual artifacts for synthetic images. Therefore, we aim to explore the capability and improve the performance of GAN methods for food image generation. Specifically, we first choose StyleGAN3 as the baseline method to generate synthetic food images and analyze the performance. Then, we identify two issues that can cause performance degradation on food images during the training phase: (1) inter-class feature entanglement during multi-food classes training and (2) loss of high-resolution detail during image downsampling. To address both issues, we propose to train one food category at a time to avoid feature entanglement and leverage image patches cropped from high-resolution datasets to retain fine details. We evaluate our method on the Food-101 dataset and show improved quality of generated synthetic food images compared with the baseline. Finally, we demonstrate the great potential of improving the performance of downstream tasks, such as food image classification by including high-quality synthetic training samples in the data augmentation.

Feature Consistency Constraints-Based CNN for Landsat Land Cover Mapping

The cascade of convolution layers and the end-to-end training process facilitate CNN feature extraction and transmission, and promote the success of CNN in image processing. However, the drawback of heavily relying on large-scale high-quality training samples restricts its applications. To avoid costly and unrealistic manual annotations for large-scale remote sensing images, existing land cover maps are considered as an alternative to manual annotations, in which noisy labels are inevitable. To alleviate the impact of noisy labels, this paper proposes to improve the consistency feature learning ability of CNNs as a feasible solution in practical land cover mapping. Firstly, an intraclass feature consistency constraint is introduced to maintain the consistency of CNN feature maps for the same class. Then, an inter-iteration feature consistency constraint is employed to guide the network to learn features that are consistent with the whole underlying distribution inside a mini-batch. These two feature consistency constraints work in a cooperative and complementary manner with the traditional cross-entropy, and together improve the consistency feature learning ability of the proposed Feature Consistency Network (FCNet). Experimental results demonstrate the effectiveness of the proposed FCNet. Extensive experiments on different network structures validate the generalization of the proposed feature consistency constraints.

Domain Adaptive Transformer Tracking Under Occlusions

Due to their excellent performance on aggregating global features, Transformer structures are being widely employed in deep learning-based visual object tracking algorithms, recently. Nevertheless, existing Transformer-based trackers still fail to handle occlusion problems due to drift in feature distributions. To address this issue, we introduce domain adaptation techniques into a novel object tracking framework, DATransT, including feature extraction, domain adaptive Transformer module and prediction head. The domain adaptive Transformer module consists of three weight-sharing branches with self and cross attention mechanisms: the source, the target and the source-target branches. Specifically, the source-target branch employs cross-attention to effectively align the feature distributions of the source and target branches. Meanwhile, we present a pseudo-labeling strategy to generate high-quality training samples. Extensive experiments show that DATransT obtains promising results on several popular datasets, containing LaSOT, TrackingNet, GOT-10k, NfS, OTB2015 and UAV123. Moreover, our method outperforms existing state-of-the-art trackers under full occlusions and partial occlusions.

Parameters tuning of multi-model database based on deep reinforcement learning

As we all know, the performance of database management system is directly linked to a vast array of knobs, which control various aspects of system operation, ranging from memory and thread counts settings to I/O optimization. Improper settings of configuration parameters are shown to have detrimental effects on performance, reliability and availability of the overall database management system. This is also true for multi-model databases, which use a single platform to support multiple data models. Existing approaches for automatic DBMS knobs tuning are not directly applicable to multi-model databases due to the diversity of multi-model database instances and workloads. Firstly, in cloud environment, they have difficulty adapting to changing environments and diverse workloads. Secondly, they rely on large-scale high-quality training samples that are difficult to obtain. Finally, they focus primarily on throughput metrics, ignoring tuning requirements for resource utilization. Therefore, in this paper, we propose a multi-model database configuration parameters tuning solution named MMDTune. It selects influential parameters, recommends the optimal configurations in a high-dimensional continuous space. For different workloads, the TD3 algorithm is improved to generate reasonable parameter adjustment plans according to the internal state of the multi-model databases. We conduct extensive experiments under 5 different workloads on real cloud databases to evaluate MMDTune. Experimental results show that MMDTune adapts well to a new hardware environment or workloads, and significantly outperforms the representative tuning tools, such as OtterTune, CDBTune.

AC-WGAN-GP: Generating Labeled Samples for Improving Hyperspectral Image Classification with Small-Samples

The lack of labeled samples severely restricts the classification performance of deep learning on hyperspectral image classification. To solve this problem, Generative Adversarial Networks (GAN) are usually used for data augmentation. However, GAN have several problems with this task, such as the poor quality of the generated samples and an unstable training process. Thereby, knowing how to construct a GAN to generate high-quality hyperspectral training samples is meaningful for the small-sample classification task of hyperspectral data. In this paper, an Auxiliary Classifier based Wasserstein GAN with Gradient Penalty (AC-WGAN-GP) was proposed. The framework includes AC-WGAN-GP, an online generation mechanism, and a sample selection algorithm. The proposed method has the following distinctive advantages. First, the input of the generator is guided by prior knowledge and a separate classifier is introduced to the architecture of AC-WGAN-GP to produce reliable labels. Second, an online generation mechanism ensures the diversity of generated samples. Third, generated samples that are similar to real data are selected. Experiments on three public hyperspectral datasets show that the generated samples follow the same distribution as the real samples and have enough diversity, which effectively expands the training set. Compared to other competitive methods, the proposed framework achieved better classification accuracy with a small number of labeled samples.

Time-series land cover mapping and urban expansion analysis using OpenStreetMap data and remote sensing big data: A case study of Guangdong-Hong Kong-Macao Greater Bay Area, China

• Optimized training samples from OpenStreetMap facilitate large-scale land cover mapping. • Sample optimization and transfer, and consistency modification strategies improve mapping results. • The proposed workflow achieves satisfactory performance and is transferable to other regions. Accurate land cover mapping provides important scientific support for ecological environment protection and sustainable urban development. However, given the high expense of acquiring training samples and the difficulty of fully utilizing remote sensing big data, large-scale time-series land cover mapping remains to be a challenge. To address the issue, we proposed a novel time-series large-scale mapping approach that obtains high-quality training samples from OpenStreetMap volunteered data and transfers them for land cover mapping in historical years. Relying on the data archived on the Google Earth Engine platform, we constructed a discriminating feature set that contains spectrum, texture, and backscatter coefficient, among others. Taking the Guangdong-Hong Kong-Macao Greater Bay Area as a case area, the annual land cover maps after spatio-temporal consistency modification from 1986 to 2021 were obtained. The validation samples proved that the derived land cover classification results have high accuracy. We verified the reliability and superiority of the proposed approach by comparing our results with existing land cover products. The proposed mapping approach owns great transferability, and the mapping results provide a valid decision-making basis for urban planning.

Sample-Efficient Deep Learning Techniques for Burn Severity Assessment with Limited Data Conditions

The automatic analysis of medical data and images to help diagnosis has recently become a major area in the application of deep learning. In general, deep learning techniques can be effective when a large high-quality dataset is available for model training. Thus, there is a need for sample-efficient learning techniques, particularly in the field of medical image analysis, as significant cost and effort are required to obtain a sufficient number of well-annotated high-quality training samples. In this paper, we address the problem of deep neural network training under sample deficiency by investigating several sample-efficient deep learning techniques. We concentrate on applying these techniques to skin burn image analysis and classification. We first build a large-scale, professionally annotated dataset of skin burn images, which enables the establishment of convolutional neural network (CNN) models for burn severity assessment with high accuracy. We then deliberately set data limitation conditions and adapt several sample-efficient techniques, such as transferable learning (TL), self-supervised learning (SSL), federated learning (FL), and generative adversarial network (GAN)-based data augmentation, to those conditions. Through comprehensive experimentation, we evaluate the sample-efficient deep learning techniques for burn severity assessment, and show, in particular, that SSL models learned on a small task-specific dataset can achieve comparable accuracy to a baseline model learned on a six-times larger dataset. We also demonstrate the applicability of FL and GANs to model training under different data limitation conditions that commonly occur in the area of healthcare and medicine where deep learning models are adopted.

Bearing Remaining Useful Life Prediction Based on AdCNN and CWGAN under Few Samples

At present, deep learning is widely used to predict the remaining useful life (RUL) of rotation machinery in failure prediction and health management (PHM). However, in the actual manufacturing process, massive rotating machinery data are not easily obtained, which will lead to the decline of the prediction accuracy of the data-driven deep learning method. Firstly, a novel prognostic framework is proposed, which is comprised of conditional Wasserstein distance-based generative adversarial networks (CWGAN) and adversarial convolution neural networks (AdCNN), which can stably generate high-quality training samples to augment the bearing degradation dataset and solve the problem of few samples. Then, the bearing RUL prediction method is realized by inputting the monitoring data into the one-dimensional convolutional neural network (1DCNN) for adversarial training. Via the bearing degradation dataset of the IEEE 2012 PHM data challenge, the reliability of the proposed method is verified. Finally, experimental results show that our approach is better than others in RUL prediction on average absolute deviation and average square root error.