Set Of Queries Research Articles

SEML: Self-Supervised Information-Enhanced Meta-learning for Few-Shot Text Classification

Training a deep-learning text classification model usually requires a large amount of labeled data, yet labeling data are usually labor-intensive and time-consuming. Few-shot text classification focuses on predicting unknown samples using only a few labeled samples. Recently, metric-based meta-learning methods have achieved promising results in few-shot text classification. They use episodic training in labeled samples to enhance the model’s generalization ability. However, existing models only focus on learning from a few labeled samples but neglect to learn from a large number of unlabeled samples. In this paper, we exploit the knowledge learned by the model in unlabeled samples to improve the generalization performance of the meta-network. Specifically, we introduce a novel knowledge distillation method that expands and enriches the meta-learning representation with self-supervised information. Meanwhile, we design a graph aggregation method that efficiently interacts the query set information with the support set information in each task and outputs a more discriminative representation. We conducted experiments on three public few-shot text classification datasets. The experimental results show that our model performs better than the state-of-the-art models in 5-way 1-shot and 5-way 5-shot cases.

SpatialFormer: Semantic and Target Aware Attentions for Few-Shot Learning

Recent Few-Shot Learning (FSL) methods put emphasis on generating a discriminative embedding features to precisely measure the similarity between support and query sets. Current CNN-based cross-attention approaches generate discriminative representations via enhancing the mutually semantic similar regions of support and query pairs. However, it suffers from two problems: CNN structure produces inaccurate attention map based on local features, and mutually similar backgrounds cause distraction. To alleviate these problems, we design a novel SpatialFormer structure to generate more accurate attention regions based on global features. Different from the traditional Transformer modeling intrinsic instance-level similarity which causes accuracy degradation in FSL, our SpatialFormer explores the semantic-level similarity between pair inputs to boost the performance. Then we derive two specific attention modules, named SpatialFormer Semantic Attention (SFSA) and SpatialFormer Target Attention (SFTA), to enhance the target object regions while reduce the background distraction. Particularly, SFSA highlights the regions with same semantic information between pair features, and SFTA finds potential foreground object regions of novel feature that are similar to base categories. Extensive experiments show that our methods are effective and achieve new state-of-the-art results on few-shot classification benchmarks.

Bi-directional Feature Reconstruction Network for Fine-Grained Few-Shot Image Classification

The main challenge for fine-grained few-shot image classification is to learn feature representations with higher inter-class and lower intra-class variations, with a mere few labelled samples. Conventional few-shot learning methods however cannot be naively adopted for this fine-grained setting -- a quick pilot study reveals that they in fact push for the opposite (i.e., lower inter-class variations and higher intra-class variations). To alleviate this problem, prior works predominately use a support set to reconstruct the query image and then utilize metric learning to determine its category. Upon careful inspection, we further reveal that such unidirectional reconstruction methods only help to increase inter-class variations and are not effective in tackling intra-class variations. In this paper, we for the first time introduce a bi-reconstruction mechanism that can simultaneously accommodate for inter-class and intra-class variations. In addition to using the support set to reconstruct the query set for increasing inter-class variations, we further use the query set to reconstruct the support set for reducing intra-class variations. This design effectively helps the model to explore more subtle and discriminative features which is key for the fine-grained problem in hand. Furthermore, we also construct a self-reconstruction module to work alongside the bi-directional module to make the features even more discriminative. Experimental results on three widely used fine-grained image classification datasets consistently show considerable improvements compared with other methods. Codes are available at: https://github.com/PRIS-CV/Bi-FRN.

Efficient Meta-Learning through Task-Specific Pseudo Labelling

Meta-learning is attracting attention as a crucial tool for few-show learning tasks. Meta-learning involves the establishment and acquisition of “meta-knowledge”, enabling the ability to adapt to a novel field using only limited data. Transductive meta-learning has garnered increasing attention as a solution to the sample bias problem arising from meta-learning’s reliance on a limited support set for adaptation. This approach surpasses the traditional inductive learning perspective, aiming to address this issue effectively. Transductive meta-learning infers the class of each instance in time by considering the relation of instances in the test set. In order to enhance the effectiveness of transductive meta-learning, this paper introduces a novel technique called task-specific pseudo labelling. The main idea is to produce synthetic labels for unannotated query sets by propagating labels from annotated support sets. This approach allows the utilization of the supervised setting as is, while incorporating the unannotated query set into the adjustment procedure. Consequently, our approach enables handling a larger number of examples during adaptation compared to inductive approaches, leading to improved classification performance of the model. Notably, this approach represents the first instance of employing task adaptation within the context of pseudo labelling. Based on the experimental outcomes in the evaluation configurations of few-shot learning, specifically in the 5-way 1-shot setup, the proposed method demonstrates noteworthy enhancements over two existing meta-learning algorithms, with improvements of 6.75% and 5.03%, respectively. Consequently, the proposed method establishes a new state-of-the-art performance in the realm of transductive meta-learning.

Computing roadmaps in unbounded smooth real algebraic sets I: Connectivity results

Answering connectivity queries in real algebraic sets is a fundamental problem in effective real algebraic geometry that finds many applications in e.g. robotics where motion planning issues are topical. This computational problem is tackled through the computation of so-called roadmaps which are real algebraic subsets of the set V under study, of dimension at most one, and which have a connected intersection with all semi-algebraically connected components of V. Algorithms for computing roadmaps rely on statements establishing connectivity properties of some well-chosen subsets of V, assuming that V is bounded.In this paper, we extend such connectivity statements by dropping the boundedness assumption on V. This exploits properties of so-called generalized polar varieties, which are critical loci of V for some well-chosen polynomial maps.

Query sampler: generating query sets for analyzing search engines using keyword research tools.

Search engine queries are the starting point for studies in different fields, such as health or political science. These studies usually aim to make statements about social phenomena. However, the queries used in the studies are often created rather unsystematically and do not correspond to actual user behavior. Therefore, the evidential value of the studies must be questioned. We address this problem by developing an approach (query sampler) to sample queries from commercial search engines, using keyword research tools designed to support search engine marketing. This allows us to generate large numbers of queries related to a given topic and derive information on how often each keyword is searched for, that is, the query volume. We empirically test our approach with queries from two published studies, and the results show that the number of queries and total search volume could be considerably expanded. Our approach has a wide range of applications for studies that seek to draw conclusions about social phenomena using search engine queries. The approach can be applied flexibly to different topics and is relatively straightforward to implement, as we provide the code for querying Google Ads API. Limitations are that the approach needs to be tested with a broader range of topics and thoroughly checked for problems with topic drift and the role of close variants provided by keyword research tools.

VLT: Vision-Language Transformer and Query Generation for Referring Segmentation.

We propose a Vision-Language Transformer (VLT) framework for referring segmentation to facilitate deep interactions among multi-modal information and enhance the holistic understanding to vision-language features. There are different ways to understand the dynamic emphasis of a language expression, especially when interacting with the image. However, the learned queries in existing transformer works are fixed after training, which cannot cope with the randomness and huge diversity of the language expressions. To address this issue, we propose a Query Generation Module, which dynamically produces multiple sets of input-specific queries to represent the diverse comprehensions of language expression. To find the best among these diverse comprehensions, so as to generate a better mask, we propose a Query Balance Module to selectively fuse the corresponding responses of the set of queries. Furthermore, to enhance the model's ability in dealing with diverse language expressions, we consider inter-sample learning to explicitly endow the model with knowledge of understanding different language expressions to the same object. We introduce masked contrastive learning to narrow down the features of different expressions for the same target object while distinguishing the features of different objects. The proposed approach is lightweight and achieves new state-of-the-art referring segmentation results consistently on five datasets.

Dual-stream coupling network with wavelet transform for cross-resolution person re-identification

Person re-identification is a prevalent technology deployed on intelligent surveillance. There have been remarkable achievements in person re-identification methods based on the assumption that all person images have a sufficiently high resolution, yet such models are not applicable to the open world. In real world, the changing distance between pedestrians and the camera renders the resolution of pedestrians captured by the camera inconsistent. When low-resolution (LR) images in the query set are matched with high-resolution (HR) images in the gallery set, it degrades the performance of the pedestrian matching task due to the absent pedestrian critical information in LR images. To address the above issues, we present a dual-stream coupling network with wavelet transform (DSCWT) for the cross-resolution person re-identification task. Firstly, we use the multi-resolution analysis principle of wavelet transform to separately process the low-frequency and high-frequency regions of LR images, which is applied to restore the lost detail information of LR images. Then, we devise a residual knowledge constrained loss function that transfers knowledge between the two streams of LR images and HR images for accessing pedestrian invariant features at various resolutions. Extensive qualitative and quantitative experiments across four benchmark datasets verify the superiority of the proposed approach.

TransCenter: Transformers With Dense Representations for Multiple-Object Tracking.

Transformers have proven superior performance for a wide variety of tasks since they were introduced. In recent years, they have drawn attention from the vision community in tasks such as image classification and object detection. Despite this wave, an accurate and efficient multiple-object tracking (MOT) method based on transformers is yet to be designed. We argue that the direct application of a transformer architecture with quadratic complexity and insufficient noise-initialized sparse queries - is not optimal for MOT. We proposeTransCenter, a transformer-based MOT architecture with dense representations for accurately tracking all the objects while keeping a reasonable runtime. Methodologically, we propose the use of image-related dense detection queries and efficient sparse tracking queries produced by our carefully designed query learning networks (QLN). On one hand, the dense image-related detection queries allow us to infer targets' locations globally and robustly through dense heatmap outputs. On the other hand, the set of sparse tracking queries efficiently interacts with image features in ourTransCenterDecoder to associate object positions through time. As a result,TransCenterexhibits remarkable performance improvements and outperforms by a large margin the current state-of-the-art methods in two standard MOT benchmarks with two tracking settings (public/private). TransCenteris also proven efficient and accurate by an extensive ablation study and, comparisons to more naive alternatives and concurrent works. The code is made publicly available at https://github.com/yihongxu/transcenter.

Conditional-based Transformer network with learnable queries for 4D deformation forecasting and tracking.

Real-time motion management for image-guided radiation therapy interventions plays an important role for accurate dose delivery. Forecasting future 4D deformations from in-plane image acquisitions is fundamental for accurate dose delivery and tumor targeting. However, anticipating visual representations is challenging and is not exempt from hurdles such as the prediction from limited dynamics, and the high-dimensionality inherent to complex deformations. Also, existing 3D tracking approaches typically need both template and search volumes as inputs, which are not available during real-time treatments. In this work, we propose an attention-based temporal prediction network where features extracted from input images are treated as tokens for the predictive task. Moreover, we employ a set of learnable queries, conditioned on prior knowledge, to predict future latent representation of deformations. Specifically, the conditioning scheme is based on estimated time-wise prior distributions computed from future images available during the training stage. Finally, we propose a new framework to address the problem of temporal 3D local tracking using cine 2D images as inputs, by employing latent vectors as gating variables to refine the motion fields over the tracked region. The tracker module is anchored on a 4D motion model, which provides both the latent vectors and the volumetric motion estimates to be refined. Our approach avoids auto-regression and leverages spatial transformations to generate the forecasted images. The tracking module reduces the error by 63% compared to a conditional-based transformer 4D motion model, yielding a mean error of 1.5±1.1 mm. Furthermore, for the studied cohort of abdominal 4D MRI images, the proposed method is able to predict future deformations with a mean geometrical error of 1.2±0.7 mm.

Transforming Primary Care Data Into the Observational Medical Outcomes Partnership Common Data Model: Development and Usability Study.

Patient-monitoring software generates a large amount of data that can be reused for clinical audits and scientific research. The Observational Health Data Sciences and Informatics (OHDSI) consortium developed the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM) to standardize electronic health record data and promote large-scale observational and longitudinal research. This study aimed to transform primary care data into the OMOP CDM format. We extracted primary care data from electronic health records at a multidisciplinary health center in Wattrelos, France. We performed structural mapping between the design of our local primary care database and the OMOP CDM tables and fields. Local French vocabularies concepts were mapped to OHDSI standard vocabularies. To validate the implementation of primary care data into the OMOP CDM format, we applied a set of queries. A practical application was achieved through the development of a dashboard. Data from 18,395 patients were implemented into the OMOP CDM, corresponding to 592,226 consultations over a period of 20 years. A total of 18 OMOP CDM tables were implemented. A total of 17 local vocabularies were identified as being related to primary care and corresponded to patient characteristics (sex, location, year of birth, and race), units of measurement, biometric measures, laboratory test results, medical histories, and drug prescriptions. During semantic mapping, 10,221 primary care concepts were mapped to standard OHDSI concepts. Five queries were used to validate the OMOP CDM by comparing the results obtained after the completion of the transformations with the results obtained in the source software. Lastly, a prototype dashboard was developed to visualize the activity of the health center, the laboratory test results, and the drug prescription data. Primary care data from a French health care facility have been implemented into the OMOP CDM format. Data concerning demographics, units, measurements, and primary care consultation steps were already available in OHDSI vocabularies. Laboratory test results and drug prescription data were mapped to available vocabularies and structured in the final model. A dashboard application provided health care professionals with feedback on their practice.

Failure analysis and prediction of roof instability in end face under repeated mining using early warning system

The overlying strata of the lower coal seam is easy to be collapsed causing the roof caving accident at the end face of the mining working face under repeated mining in close-distance coal seams. In order to predict the roof instability of the end face, the mechanical model of the granular arch structure is established in this study to further analyze its main influencing factors. The results show that the mining height of the working face, the advancing speed, the distance of coal seams, the tip-to-face distance, the strength of the surrounding rock and the support setting the load of the support are the main influencing factors on the roof caving of the end face. Subsequently, the prediction model of roof instability in the end face under repeated mining is constructed through the radial basis function neural network (RBFNN) and the above main influencing factors are regarded as input layer indexes. Meanwhile, the roof subsidence, coal wall deformation and support load are determined as the output layer indexes. The predicted results are closer to the results of sample tests. Finally, the early warning system, including monitoring and early warning, data query, emergency management, user management, and system settings, is designed to monitor roof conditions of the end face and timely warn the roof accidents. The field application proves that the system has good practical value, which is of great significance to intelligent prediction of coal mine stope disaster and prevent the end face roof disaster under repeated mining and. This will promote the safe and efficient construction of coal mine production.

Using CORONA Imagery to Study Land Use and Land Cover Change—A Review of Applications

CORONA spy satellites offer high spatial resolution imagery acquired in the 1960s and early 1970s and declassified in 1995, and they have been used in various scientific fields, such as archaeology, geomorphology, geology, and land change research. The images are panchromatic but contain many details of objects on the land surface due to their high spatial resolution. This systematic review aims to study the use of CORONA imagery in land use and land cover change (LULC) research. Based on a set of queries conducted on the SCOPUS database, we identified and examined 54 research papers using such data in their study of LULC. Our analysis considered case-study area distributions, LULC classes and LULC changes, as well as the methods and types of geospatial data used alongside CORONA data. While the use of CORONA images has increased over time, their potential has not been fully explored due to difficulties in processing CORONA images. In most cases, study areas are small and below 5000 km2 because of the reported drawbacks related to data acquisition frequency, data quality and analysis. While CORONA imagery allows analyzing built-up areas, infrastructure and individual buildings due to its high spatial resolution and initial mission design, in LULC studies, researchers use the data mostly to study forests. In most case studies, CORONA imagery was used to extend the study period into the 1960s, with only some examples of using CORONA alongside older historical data. Our analysis proves that in order to detect LULC changes, CORONA can be compared with various contemporary geospatial data, particularly high and very high-resolution satellite imagery, as well as aerial imagery.

Joint upper & expected value normalization for evaluation of retrieval systems: A case study with Learning-to-Rank methods

While original IR evaluation metrics are normalized in terms of their upper bounds based on an ideal ranked list, a corresponding expected value normalization for them has not yet been studied. We present a framework with both upper and expected value normalization, where the expected value is estimated from a randomized ranking of the corresponding documents present in the evaluation set. We next conducted two case studies by instantiating the new framework for two popular IR evaluation metrics (e.g., nDCG, MAP) and then comparing them against the traditional metrics.Experiments on two Learning-to-Rank (LETOR) benchmark data sets, MSLR-WEB30K (includes 30K queries and 3771K documents) and MQ2007 (includes 1700 queries and 60K documents), with eight LETOR methods (pairwise & listwise), demonstrate the following properties of the new expected value normalized metric: (1) Statistically significant differences (between two methods) in terms of original metric no longer remain statistically significant in terms of Upper Expected(UE) normalized version and vice-versa, especially for uninformative query-sets. (2) When compared against the original metric, our proposed UE normalized metrics demonstrate an average of 23% and 19% increase in terms of Discriminatory Power on MSLR-WEB30K and MQ2007 data sets, respectively. We found similar improvements in terms of consistency as well; for example, UE-normalized MAP decreases the swap rate by 28% while comparing across different data sets and 26% across different query sets within the same data set. These findings suggest that the IR community should consider UE normalization seriously when computing nDCG and MAP and more in-depth study of UE normalization for general IR evaluation is warranted.

Can ChatGPT Help in Electronics Research and Development? A Case Study with Applied Sensors.

In this paper, we investigated the applicability of ChatGPT AI in electronics research and development via a case study of applied sensors in embedded electronic systems, a topic that is rarely mentioned in the recent literature, thus providing new insight for professionals and academics. The initial electronics-development tasks of a smart home project were prompted to the ChatGPT system to find out its capabilities and limitations. We wanted to obtain detailed information on the central processing controller units and the actual sensors usable for the specific project, their specifications and recommendations on the hardware and software design flow additionally. Furthermore, an extensive literature survey was requested to see if the bot could offer scientific papers covering the given topic. It was found that the ChatGPT responded with proper recommendations on controllers. However, the suggested sensor units, the hardware and software design were only partially acceptable, with occasional errors in specifications and generated code. The results of the literature survey showed that non-acceptable, fabricated citations (fake authors list, title, journal details and DOI-Digital Object identifier) were presented by the bot. The paper provides a detailed qualitative analysis, a performance analysis and critical discussion of the aforementioned aspects while providing the query set, the generated answers and codes as supplied data with the goal to give added value to electronics researchers and developers if trying to reach out for the tools in their profession.

Recommending tasks based on search queries and missions

Abstract Web search is an experience that naturally lends itself to recommendations, including query suggestions and related entities. In this article, we propose to recommend specific tasks to users, based on their search queries, such as planning a holiday trip or organizing a party. Specifically, we introduce the problem of query-based task recommendation and develop methods that combine well-established term-based ranking techniques with continuous semantic representations, including sentence representations from several transformer-based models. Using a purpose-built test collection, we find that our method is able to significantly outperform a strong text-based baseline. Further, we extend our approach to using a set of queries that all share the same underlying task, referred to as search mission, as input. The study is rounded off with a detailed feature and query analysis.

A Few-Shot Defect Detection Method for Transmission Lines Based on Meta-Attention and Feature Reconstruction

In tasks of transmission line defect detection, traditional object detection algorithms are ineffective, with few training samples of defective components. Meta-learning uses multi-task learning as well as fine-tuning to learn common features in different tasks, which has the ability to adapt to new tasks quickly, shows good performance in few-shot object detection, and has good generalization in new tasks. For this reason, we proposed a few-shot defect detection method (Meta PowerNet) with a Meta-attention RPN and Feature Reconstruction Module for transmission lines based on meta-learning. First, in the stage of region proposal, a new region proposal network (Meta-Attention Region Proposal Network, MA-RPN) is designed to fuse the support set features and the query set features to filter the noise in anchor boxes. In addition, it has the ability to focus on the subtle texture features of smaller-sized objects by fusing low-level features from the query set. Second, in the meta-feature construction stage, we designed a meta-learner with the defect feature reconstruction module as the core to capture and focus on the defect-related feature channels. The experimental results show that under the condition, there are only 30 training objects for various types of component defects. The method achieves 72.5% detection accuracy for component defects, which is a significant improvement compared with other mainstream few-shot object detection. Meanwhile, the MA-RPN designed in this paper can be used in other meta-learning object detection models universally.

Observational Equivalence of Conditional Belief Bases

In nonmonotonic reasoning, a conditional of the form ‘If A then usually B’ is typically accepted if a situation where both A and B hold is deemed to be more plausible, more probable, or less surprising, etc., than a situation where A holds, but B does not hold. In a propositional setting, this leads to a relation on the propositional interpretations, also called worlds, by comparing worlds according to their plausibility. In this paper, we address the question of which kind of relations on the set of worlds can be obtained by completing a conditional belief base via an inductive inference operator. As a key concept for our investigations, we introduce and employ the notion of observational equivalence of belief bases that takes an inductive inference operator and a set of queries into account. This leads to the notion of the observational normal form (ONF) and, by focussing on so-called base conditionals, to the base conditional normal form (BCNF). The acceptance of base conditionals corresponds to the plausibility ordering on possible worlds induced by an inference method. Both normal forms ONF and BCNF are combined with renamings as an additional dimension. We establish the interrelationships among the normal forms and evaluate them empirically with respect to systematically generated belief bases.

Capsule neural tensor networks with multi-aspect information for Few-shot Knowledge Graph Completion

Few-shot Knowledge Graph Completion (FKGC) has recently attracted significant research interest due to its ability to expand few-shot relation coverage in Knowledge Graphs. Prevailing FKGC approaches focus on exploiting the one-hop neighbor information of entities to enhance few-shot relation embedding. However, these methods select one-hop neighbors randomly and neglect the rich multi-aspect information of entities. Although some methods have attempted to leverage Long Short-Term Memory (LSTM) to learn few-shot relation embedding, they are sensitive to the input order. To address these limitations, we propose the Capsule Neural Tensor Networks with Multi-Aspect Information approach (short for InforMix-FKGC). InforMix-FKGC employs a one-hop neighbor selection strategy based on how valuable they are and encodes multi-aspect information of entities, including one-hop neighbors, attributes and literal description. Then, a capsule network is responsible for integrating the support set and deriving few-shot relation embedding. Moreover, a neural tensor network is used to match the query set with the support set. In this way, InforMix-FKGC can learn few-shot relation embedding more precisely so as to enhance the accuracy of FKGC. Extensive experiments on the NELL-One and Wiki-One datasets demonstrate that InforMix-FKGC significantly outperforms ten state-of-the-art methods in terms of Mean Reciprocal Rank and Hits@K.

How the Quality of Maintenance Tasks is Affected by Criteria for Selecting Engineers for Collaboration

In industry, software projects might span over decades, with many engineers joining or leaving the company over time. In these circumstances, no single engineer has all of the knowledge when maintenance tasks such as Traceability Link Recovery (TLR), Bug Localization (BL), and Feature Location (FL) are performed. Thus, collaboration has the potential to boost the quality of maintenance tasks since the solution advanced by one engineer might be enhanced with contributions from other engineers. However, assembling a team of software engineers to collaborate may not be as intuitive as we might think. In the context of a worldwide industrial supplier of railway solutions, this work evaluates how the quality of TLR, BL, and FL is affected by the criteria for selecting engineers for collaboration. The criteria for collaboration are based on engineers’ profile information to select the set of search queries that are involved in the maintenance task. Collaboration is achieved by applying automatic query reformulation, and the location relies on an evolutionary algorithm. Our work uncovers how software engineers who might be seen as not being relevant in the collaboration can lead to significantly better results. A focus group confirmed the relevance of the findings.