Multi-objective optimization (MOO) has become an important method in machine learning, which involves solving multiple competing objective problems simultaneously. Nowadays, many MOO algorithms assume that gradient information is easily available and use this to optimize functions. However, when encountering situations where gradients are not available, such as black-box functions or non-differentiable functions, these ineffective. In paper, we propose a zeroth-order algorithm named SZMG...

In the domain of natural language processing (NLP), a primary challenge pertains to process Chinese tokenization, which remains challenging due lack explicit word boundaries in written Chinese. The existing tokenization methods often treat each character as an indivisible unit, neglecting finer semantic features embedded characters, such radicals. To tackle this issue, we propose novel token representation method that integrates radical-based into process. proposed extends vocabulary include...

Abstract With the increasing of requirements from many aspects, various queries and analyses arise focusing on social network. Time is a common necessary dimension in types networks. Social networks with time information are called temporal networks, which can be when user sends message to another user. Keywords search consists finding relationships between group users that has set query labels valid within interval. It provides assistance network analysis, classification users, community...

Crowd flows prediction is an important problem of urban computing. The existing best-known method adopts deep residual networks to model spatio-temporal properties and often achieves good performance. However, since three separated network structures are used the properties, time cost expensive for method. In this paper, we propose improved reduce running by simplifying its architecture. Compared with method, training predicting our can be reduced dramatically. Moreover, achieve similar...

Top-k nearest keyword search is important for various applications. However, the existing methods are only applicable to static graphs, not public transportation networks. This because unlike graph, network a temporal graph where path in must satisfy time constraint. Thus, which reachable may graph. Therefore, graphs cannot be applied graphs. In this paper, solve top-k neighbor on networks, we propose two indexes and algorithms called Temporal Forward Search (TFS) Forward-Backward (TFBS)...

In this paper, we consider the path queries in public transportation networks which are widely used planning and location-based services. general, index is often adopted to speed up networks. However, pre-processing for building existing best-known query algorithms not efficient. Aimed at shortcoming, propose an improved index, based on MapReduce queries. Specifically, proposed algorithm includes map reduce phases. phase, network divided into different subgraphs generate key-value pairs of...

Text deduplication is an important operation for text document analysis applications. Given a set of documents, we often need to remove the documents whose similarity values are not less than specified threshold. However, if similar be removed too large, remaining may enough analysis. In this paper, consider problem on how balance and documents. We try reduce duplication information as much possible with minimum number removed. propose greedy algorithm our based concept graph which can...

A new kind of multifunctional energy-saving heave compensation winch(HCW) for ultra-depth floating drilling, which has functions lifting, and energy-recycle, is presented based on direct drive volume control(DDVC) hydraulic transformer(HT) energy-recycle technology. The simulation model HCW built by using AMESim software, the structure main controller designed, consist a controller, HT pump source(DDPS) controller. flowrate calculation method distribution algorithm are studied, flow...

The computation of a group Steiner tree (GST) in various types graph networks, such as social network and transportation network, is fundamental problem graphs, with important applications. In these time common necessary dimension, for example, information can be the when user sends message to another user. Graphs called temporal graphs. However, few studies have been conducted on GST terms This study analyzes i.e., (TGST), which shown an NP-hard problem. We propose efficient solution based...

In this paper, we consider the top-k route search with user's preferences. Specifically, given a set of POIs, our problem is to find k different routes from source POI target such that constraint on cost and POIs covered by can optimally satisfy user-defined weighted feature preference. It has been shown NP-hard. The challenge how select plenty construct an optimal especially when size candidate large. order support large dataset or looser budget constraint, propose parallel method single...

Abstract Keyword search on temporal graph is to find a tree covering set of query labels and being valid in the time interval. It has many applications cloud computing, community detection, social network, collaborative project, so on. However, existing methods are limited solving problem keyword graphs. We propose two basic algorithms, discrete timestamp algorithm, approximate algorithm , idea which trying turn into traditional graph. To address low efficiency quality we new based dynamic...