Previous chapter Next Full AccessProceedings Proceedings of the 2013 SIAM International Conference on Data Mining (SDM)Multi-View Clustering via Joint Nonnegative Matrix FactorizationJialu Liu, Chi Wang, Jing Gao, and Jiawei HanJialu Hanpp.252 - 260Chapter DOI:https://doi.org/10.1137/1.9781611972832.28PDFBibTexSections ToolsAdd to favoritesExport CitationTrack CitationsEmail SectionsAboutAbstract Many real-world datasets are comprised different representations or views which often provide...

While most topic modeling algorithms model text corpora with unigrams, human interpretation often relies on inherent grouping of terms into phrases. As such, we consider the problem discovering topical phrases mixed lengths. Existing work either performs post processing to results unigram-based models, or utilizes complex n-gram-discovery models. These methods generally produce low-quality suffer from poor scalability even moderately-sized datasets. We propose a different approach that is...

Text data are ubiquitous and play an essential role in big applications. However, text mostly unstructured. Transforming unstructured into structured units (e.g., semantically meaningful phrases) will substantially reduce semantic ambiguity enhance the power efficiency at manipulating such using database technology. Thus mining quality phrases is a critical research problem field of databases. In this paper, we propose new framework that extracts from corpora integrated with phrasal...

Query optimizers depend on selectivity estimates of query predicates to produce a good execution plan. When contains multiple predicates, today's use variety assumptions, such as independence between estimate selectivity. While techniques have the benefit fast estimation and small memory footprint, they often incur large errors. In this work, we reconsider regression problem. We explore application neural networks tree-based ensembles important problem multi-dimensional range predicates....

AutoGen is an open-source framework that allows developers to build LLM applications via multiple agents can converse with each other accomplish tasks. are customizable, conversable, and operate in various modes employ combinations of LLMs, human inputs, tools. Using AutoGen, also flexibly define agent interaction behaviors. Both natural language computer code be used program flexible conversation patterns for different applications. serves as a generic infrastructure diverse complexities...

The sensation of being able to feel the shape an object when grasping it in Virtual Reality (VR) enhances a sense presence and ease manipulation. Though most prior works focus on force feedback fingers, haptic emulation 3D requires touch using entire hand. Hence, we present Pop-up Prop Palm (PuPoP), light-weight pneumatic shape-proxy interface worn palm that pops several airbags up with predefined primitive shapes for grasping. When user's hand encounters virtual object, airbag appropriate...

This paper proposes a novel active boundary loss for semantic segmentation. It can progressively encourage the alignment between predicted boundaries and ground-truth during end-to-end training, which is not explicitly enforced in commonly used cross-entropy loss. Based on detected from segmentation results using current network parameters, we formulate problem as differentiable direction vector prediction to guide movement of each iteration. Our model-agnostic be plugged training networks...

Content and style disentanglement is an effective way to achieve few-shot font generation. It allows transfer the of image in a source domain defined with few reference images target domain. However, content feature extracted using representative might not be optimal. In light this, we propose fusion module (CFM) project into linear space by features basis fonts, which can take variation caused different fonts consideration. Our method also optimize representation vector through lightweight...

Entity recognition is an important but challenging research problem. In reality, many text collections are from specific, dynamic, or emerging domains, which poses significant new challenges for entity with increase in name ambiguity and context sparsity, requiring detection without domain restriction. this paper, we investigate (ER) distant-supervision propose a novel relation phrase-based ER framework, called ClusType, that runs data-driven phrase mining to generate mention candidates...

Analysts need interactive speed for exploratory analysis, but big data systems are often slow. With sampling, can produce approximate answers fast enough visualization, at the cost of accuracy and trust. We propose optimistic which approaches these issues from a user experience perspective. This method lets analysts explore results interactively, provides way to detect recover errors later. Pangloss implements ideas. discuss design raised by visualization systems. test this concept with five...

Negative curvature in the relatively new Ollivier-Ricci sense of a wireless network graph is shown to be culprit behind large queue occupancy, routing energy, and restricted capacity region under any throughput optimal protocol. This counterpart congestion phenomenon occurring Gromov negatively curved wired least cost path routing. Significantly different protocols call for significantly concepts explain "congestion" phenomenon. The rationale that it defined terms transportation...

In this letter, we proposed a novel band selection algorithm for hyperspectral images (HSIs) based on column subset selection. The main idea of the comes from problem in numerical linear algebra. It selects group bands, which maximizes volume selected columns. Since high dimensionality decreases contrast between use Manhattan distance to obtain higher quality. Experimental results real HSIs show that obtains competitively good results, terms classification accuracy, and is robust noisy bands.

Abstract Background Malignant liver tumor is one of the main causes human death. In order to help physician better diagnose and make personalized treatment schemes, in clinical practice, it often necessary segment visualize from abdominal computed tomography images. Due large number slices sequence, developing an automatic reliable segmentation method very favored by physicians. However, because noise existed scan sequence similar pixel intensity tumors with their surrounding tissues,...

Object detection is an important part of autonomous driving technology. To ensure the safe running vehicles at high speed, real‐time and accurate all objects on road required. How to balance speed accuracy a hot research topic in recent years. This paper puts forward one‐stage object algorithm based YOLOv4, which improves supports operation. The backbone doubles stacking times last residual block CSPDarkNet53. neck replaces SPP with RFB structure, PAN structure feature fusion module, adds...

The traditional interferometric calibration of phase spatial light modulators (SLM) based on interference fringes shift is easily disturbed due to environmental vibration. Here a kind absolutely SLM investigated eliminate the disturbance using dual honeycomb gratings composited with Billet-split Fresnel zone plates (BsFZP), in which split an incident beam into three beams and first two are interfered by BsFZP while last chosen as absolute reference point. experiments both 532 632.8 nm...

Urban climate models are critical for understanding and addressing the impacts of urban change. Yet, process-based face limitations high-entry barriers substantial computing resource consumption, prompting development data-driven methods. However, recently developed emulators, being location-dependent, less scalable may overlook geospatial data. In this study, we develop location-independent machine learning emulators daily maximum canyon air temperature. To overcome complexities associated...

Probe interval graphs have been introduced in the physical mapping and sequencing of DNA as a generalization graphs. We prove that probe are weakly triangulated, hence perfect, characterize by consecutive orders their intrinsic cliques.

Taxonomies consist of machine-interpretable semantics and provide valuable knowledge for many web applications. For example, online retailers (e.g., Amazon eBay) use taxonomies product recommendation, search engines Google Bing) leverage to enhance query understanding. Enormous efforts have been made on constructing either manually or semi-automatically. However, with the fast-growing volume content, existing will become outdated fail capture emerging knowledge. Therefore, in applications,...

This paper proposes a novel deep learning-based video object matting method that can achieve temporally coherent results. Its key component is an attention-based temporal aggregation module maximizes image networks' strength for networks. computes correlations pixels adjacent to each other along the time axis in feature space, which robust against motion noises. We also design loss term train attention weights, drastically boosts performance. Besides, we show how effectively solve trimap...