Nearest neighbor search aims at obtaining the samples in database with smallest distances from them to queries, which is a basic task range of fields, including computer vision and data mining. Hashing one most widely used methods for its computational storage efficiency. With development deep learning, hashing show more advantages than traditional methods. In this survey, we detailedly investigate current algorithms supervised unsupervised hashing. Specifically, categorize into pairwise...

Recently, some contrastive learning methods have been proposed to simultaneously learn representations and clustering assignments, achieving significant improvements. However, these do not take the category information objective into consideration, thus learned are optimal for performance might be limited. Towards this issue, we first propose a novel graph framework, then apply it task, resulting in Graph Constrastive Clustering (GCC) method. Different from basic that only assumes an image...

Knowledge graph (KG) contains well-structured external information and has shown to be effective for high-quality recommendation. However, existing KG enhanced recommendation methods have largely focused on exploring advanced neural network architectures better investigate the structural of KG. While model learning, these mainly rely Negative Sampling (NS) optimize models both embedding task task. Since NS is not robust (e.g., sampling a small fraction negative instances may lose lots useful...

This article studies self-supervised graph representation learning, which is critical to various tasks, such as protein property prediction. Existing methods typically aggregate representations of each individual node representations, but fail comprehensively explore local substructures (i.e., motifs and subgraphs), also play important roles in many mining tasks. In this article, we propose a learning framework named cluster-enhanced Contrast (CLEAR) that models the structural semantics from...

In this paper, we study semi-supervised graph classification, a fundamental problem in data mining and machine learning. The is typically solved by learning neural networks with pseudo-labeling or knowledge distillation to incorporate both labeled unlabeled graphs. However, these methods usually either suffer from overconfident biased pseudo-labels suboptimal caused the insufficient use of data. Inspired recent progress contrastive dual learning, propose DualGraph, principled framework...

This paper studies the problem of traffic flow forecasting, which aims to predict future conditions on basis road networks and in past. The is typically solved by modeling complex spatio-temporal correlations data using graph neural (GNNs). However, performance these methods still far from satisfactory since GNNs usually have limited representation capacity when it comes networks. Graphs, nature, fall short capturing non-pairwise relations. Even worse, existing follow paradigm message...

Abstract Motivation Single cell RNA-sequencing (scRNA-seq) technology enables whole transcriptome profiling at single resolution and holds great promises in many biological medical applications. Nevertheless, scRNA-seq often fails to capture expressed genes, leading the prominent dropout problem. These dropouts cause problems down-stream analysis, such as significant increase of noises, power loss differential expression analysis obscuring gene-to-gene or cell-to-cell relationship....

This paper studies semi-supervised graph classification, a crucial task with wide range of applications in social network analysis and bioinformatics. Recent works typically adopt neural networks to learn graph-level representations for failing explicitly leverage features derived from topology (e.g., paths). Moreover, when labeled data is scarce, these methods are far satisfactory due their insufficient exploration unlabeled data. We address the challenge by proposing novel framework called...

Graph classification is a critical task in numerous multimedia applications, where graphs are employed to represent diverse types of data, including images, videos, and social networks. Nevertheless, the real world, labeled graph data always limited or scarce. To address this issue, we focus on semi-supervised task, which involves both supervised unsupervised models learning from unlabeled data. In contrast recent approaches that transfer entire knowledge model one, argue an effective should...

Recently, many unsupervised deep learning methods have been proposed to learn clustering with unlabelled data. By introducing data augmentation, most of the latest look into from perspective that original image and its transformation should share similar semantic assignment. However, representation features could be quite different even they are assigned same cluster since softmax function is only sensitive maximum value. This may result in high intra-class diversities feature space, which...

The rapidly developing quantum technologies and thermodynamics have put forward a requirement to precisely control measure the temperature of microscopic matter at level. Many thermometry schemes been proposed. However, measuring low is still challenging because obtained sensing errors generally tend diverge with decreasing temperature. Using continuous-variable system as thermometer, we propose non-Markovian reservoir. A mechanism make error $\ensuremath{\delta}T$ scale $T$ Landau bound...

Graph neural networks (GNNs) have emerged as powerful tools for graph classification tasks. However, contemporary methods are predominantly studied in fully supervised scenarios, while there could be label ambiguity and noise real-world applications. In this work, we explore the weakly problem of partial learning on graphs, where each sample is assigned a collection candidate labels. A novel method called <underline xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Controlling the decoherence induced by interaction of quantum system with its environment is a fundamental challenge in technology. Utilizing Floquet theory, we explore constructive role temporal periodic driving suppressing spin-1/2 particle coupled to spin bath. It revealed that, accompanying formation bound state quasienergy spectrum whole including and environment, dissipation can be inhibited tends coherently synchronize driving. seen as an analog suppression structured spatially...

Hypergraph Convolutional Network (HCN) has be-come a proper choice for capturing high-order relationships. Existing HCN methods are tailored static hypergraphs, which unsuitable the dynamic evolution in real-world scenarios. In this paper, we explore based on attention mechanism (DyHCN) time series prediction. It not only effectively exploits spatial and temporal relationships hypergraph, but also continuously aggregates cues of time-varying hypergraphs with global local embeddings....

Particle filters have been introduced as a powerful tool to estimate the posterior density of nonlinear systems. These are also capable processing data online required in many practical applications. In this paper, we propose novel technique for video stabilization based on particle filtering framework. Scale-invariant feature points extracted form rough which is used model importance density. We use constant-velocity Kalman filter intentional camera movement. prove that will lower error...

Nearest neighbor search aims to obtain the samples in database with smallest distances from them queries, which is a basic task range of fields, including computer vision and data mining. Hashing one most widely used methods for its computational storage efficiency. With development deep learning, hashing show more advantages than traditional methods. In this survey, we detailedly investigate current algorithms supervised unsupervised hashing. Specifically, categorize into pairwise methods,...

Knowledge Graph (KG) is a flexible structure that able to describe the complex relationship between data entities. Currently, most KG embedding models are trained based on negative sampling, i.e., model aims maximize some similarity of connected entities in KG, while minimizing sampled disconnected Negative sampling helps reduce time complexity learning by only considering subset instances, which may fail deliver stable performance due uncertainty procedure. To avoid such deficiency, we...

Recently, hashing is widely used in approximate nearest neighbor search for its storage and computational efficiency. Most of the unsupervised methods learn to map images into semantic similarity-preserving hash codes by constructing local similarity structure from pre-trained model as guiding information, i.e., treating each point pair similar if their distance small feature space. However, due inefficient representation ability model, many false positives negatives will be introduced lead...

The majority of deep unsupervised hashing methods usually first construct pairwise semantic similarity information and then learn to map images into compact hash codes while preserving the structure, which implies that quality highly depends on constructed structure. However, since features for each kind semantics scatter in high-dimensional space with unknown distribution, previous could introduce a large number false positives negatives boundary points distributions local structure based...

This paper delves into the problem of correlated time-series forecasting in practical applications, an area growing interest a multitude fields such as stock price prediction and traffic demand analysis. Current methodologies primarily represent data using conventional graph structures, yet these fail to capture intricate structures with non-pairwise relationships. To address this challenge, we adopt dynamic hypergraphs study better illustrate complex interactions, introduce novel hypergraph...