With the existence of many large transaction databases, huge amounts data, high scalability distributed systems, and easy partitioning distribution a centralized database, it is important to investigate efficient methods for mining association rules. The study discloses some interesting relationships between locally globally item sets proposes an rule algorithm, FDM (fast rules), which generates small number candidate substantially reduces messages be passed at A performance shows that has...

Many sequential algorithms have been proposed for the mining of association rules. However, very little work has done in rules distributed databases. A direct application to databases is not effective, because it requires a large amount communication overhead. In this study, an efficient algorithm called DMA (Distributed Mining Association rules), proposed. It generates small number candidate sets and only O(n) messages support-count exchange each set, where n sites database. The implemented...

A top-down progressive deepening method is developed for efficient mining of multiple-level association rules from large transaction databases based on the a priori principle. group variant algorithms proposed ways sharing intermediate results, with relative performance tested and analyzed. The enforcement different interestingness measurements to find more interesting rules, relaxation rule conditions finding "level-crossing" are also investigated. study shows that can be discovery strong rules.

With the memory-resource-limited constraints, class-incremental learning (CIL) usually suffers from "catastrophic forgetting" problem when updating joint classification model on arrival of newly added classes. To cope with forgetting problem, many CIL methods transfer knowledge old classes by preserving some exemplar samples into size-constrained memory buffer. utilize buffer more efficiently, we propose to keep auxiliary low-fidelity samples, rather than original real-high-fidelity samples....

As a challenging problem, few-shot class-incremental learning (FSCIL) continually learns sequence of tasks, confronting the dilemma between slow forgetting old knowledge and fast adaptation to new knowledge. In this paper, we concentrate on "slow versus fast" (SvF) determine which components be updated in fashion or fashion, thereby balance old-knowledge preservation new-knowledge adaptation. We propose multi-grained SvF strategy cope with from two different grains: intra-space (within same...

Knowledge discovery facilitates querying database knowledge and intelligent query answering in systems. We investigate the application of discovered knowledge, concept hierarchies, tools for A knowledge-rich data model is constructed to incorporate tools. Queries are classified into queries queries. Both types can be answered directly by simple retrieval or intelligently analyzing intent providing generalized, neighborhood associated information using stored knowledge. Techniques have been...

Different from many other attributes, facial expression can change in a continuous way, and therefore, slight semantic of input should also lead to the output fluctuation limited small scale. This consistency is important. However, current Facial Expression Recognition (FER) datasets may have extreme imbalance problem, as well lack data excessive amounts noise, hindering this leading performance decreasing when testing. In paper, we not only consider prediction accuracy on sample points, but...

Real-time understanding of surrounding environment is an essential yet challenging task for autonomous driving system. The system must not only deliver accurate result but also low latency performance. In this paper, we focus on the fast-and-accurate semantic segmentation. An efficient and powerful deep neural network termed as Driving Segmentation Network (DSNet) a novel loss function Object Weighted Focal Loss are proposed. designing DSNet, our goal to achieve best capacity with...

With the rapid development of social media, tremendous videos with new classes are generated daily, which raise an urgent demand for video classification methods that can continuously update while maintaining knowledge old limited storage and computing resources. In this paper, we summarize task as Class-Incremental Video Classification (CIVC) propose a novel framework to address it. As subarea incremental learning tasks, challenge catastrophic forgetting is unavoidable in CIVC. To better...

Different from universal object detection, referring expression comprehension (REC) aims to locate specific objects referred by natural language expressions. The provides high-level concepts of relevant visual and contextual patterns, which vary significantly with different expressions account for only a few those encoded in the REC model. This leads us question: do we really need entire network fixed structure various expressions? Ideally, given an expression, expression-relevant components...

As an important and challenging problem, gait recognition has gained considerable attention. It suffers from confounding conditions, that is, it is sensitive to camera views, dressing types so on. Interestingly, observed that, under different local body parts contribute differently performance. In this paper, we propose a condition-aware comparison scheme measure pairs' similarity via novel module named Instructor. Also, present geometry-guided data augmentation approach (Dresser) enrich...

Abstract In this paper, an approach for reorganizing Web sites based on user access patterns is proposed. Our goal to build adaptive by evolving site structure facilitate access. The consists of three steps: preprocessing, page classification, and reorganization. pages a are processed create internal representation the site. Page information its users extracted from server log. classified into two categories, index content pages, information. After classified, in reorganization, examined...

Silent Speech Interface (SSI) has been proposed as a means of reconstructing audible speech from silent articulatory gestures for covert voice communication in public and assistance the aphasic. Prior arts SSI, either relying on wearable devices or cameras, may lead to extended contact requirements privacy leakage risks. The recent advances acoustic sensing have brought new opportunities gestures, but their original intention is infer content classification instead reconstruction, resulting...

As a challenging problem, few-shot class-incremental learning (FSCIL) continually learns sequence of tasks, confronting the dilemma between slow forgetting old knowledge and fast adaptation to new knowledge. In this paper, we concentrate on "slow vs. fast" (SvF) determine which components be updated in fashion or fashion, thereby balance old-knowledge preservation new-knowledge adaptation. We propose multi-grained SvF strategy cope with from two different grains: intra-space (within same...