Author name cluster

Xiaoyu He

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

7 papers

2 author rows

EAAI Journal 2025 Journal Article

A three-dimensional dynamic spatial-temporal graph neural network for ocean temperature field prediction

Shuai Zhang
ZhuoLin Li
Xiaoyu He
Jie Yu
LingYu Xu

Accurate prediction of the ocean temperature field is vital for the protection of marine ecosystems under climate change. However, most existing methods only consider temporal changes, ignoring the rich three-dimensional (3D) dynamic spatial characteristics of the ocean temperature field. To improve prediction accuracy, we propose a fine-grained modeling method that uses the spatial correlations in the ocean temperature field, called the 3D Dynamic Spatial-Temporal Graph Neural Network (3D-DSTGN). Specifically, according to the 3D spatial features and dynamic spatial dependencies of the ocean temperature field, we first decompose the spatial correlation of the ocean temperature field into long-term static and short-term dynamic parts through statistical analysis of real ocean temperature datasets. We then build a 3D dynamic graph structure learning module to create static and dynamic graph structures with 3D spatial features to model and capture the corresponding spatial correlations. Next, based on the two graph structures, we apply a dual-mode graph convolution block to fully capture the dynamic spatial dependencies of the ocean temperature field. Furthermore, we use a multi-scale temporal convolution block to capture complex temporal dependencies from historical ocean temperature data. Finally, the dual-mode graph convolution block and the multi-scale temporal convolution block construct the spatio-temporal recurrent module, which extracts complex dynamic spatio-temporal contextual dependencies. On a large-scale real ocean temperature dataset from the sea surface to a subsurface depth of approximately 2, 000 m, 3D-DSTGN outperforms other baselines in experiments across different temporal and spatial scales, effectively modeling the dynamic three-dimensional spatial relationships of the ocean temperature field.

Details DOI

AIJ Journal 2024 Journal Article

A stochastic process approach for multi-agent path finding with non-asymptotic performance guarantees

Xiaoyu He
Xueyan Tang
Wentong Cai
Jingning Li

Multi-agent path finding (MAPF) is a classical NP-hard problem that considers planning collision-free paths for multiple agents simultaneously. A MAPF problem is typically solved via addressing a sequence of single-agent path finding subproblems in which well-studied algorithms such as A ⁎ are applicable. Existing methods based on this idea, however, rely on an exhaustive search and therefore only have asymptotic performance guarantees. In this article, we provide a modeling paradigm that converts a MAPF problem into a stochastic process and adopts a confidence bound based rule for finding the optimal state transition strategy. A randomized algorithm is proposed to solve this stochastic process, which combines ideas from conflict based search and Monte Carlo tree search. We show that the proposed method is almost surely optimal while enjoying non-asymptotic performance guarantees. In particular, the proposed method can, after solving N single-agent subproblems, produce a feasible solution with suboptimality bounded by O ( 1 / N ). The theoretical results are verified by several numerical experiments based on grid maps.

Details DOI

EAAI Journal 2024 Journal Article

MTFR: An universal multimodal fusion method through Modality Transfer and Fusion Refinement

Xueyu Guo
Shengwei Tian
Long Yu
Xiaoyu He
Zhenbo Wang

Multimodal learning has made great achievements in recent years. However, most of the existing modal fusion methods capture cross-modal correlations in a single stage and are designed for specific tasks, with poor generalizability and portability. In this work, we propose the Modality Transfer and Fusion Refinement method (MTFR), which contains two feature-level fusion modules and can be used for several applications. Specifically, the Modality Transfer greatly shortens the distance between modalities through transferring non-text information into text-like representations. Fusion Refinement corrects and complements the previous fusion process to make the model focus more on task-specific knowledge by introducing uni-modal processing. To illustrate the validity of MTFR, we design three experiments for sarcasm detection, sentiment analysis, and offensiveness detection. The experimental results validate that our method achieves competitive performance in all three applications and outperforms both task-specific models and multimodal fusion baselines, which demonstrate the good generalizability and robustness of our method. Thanks to the promising performance, the method will facilitate the development of multimodal fusion and can promote multimodal applications in practice such as social sentiment analysis and multimodal conversational robots.

Details DOI

STOC Conference 2023 Conference Paper

Approximating Binary Longest Common Subsequence in Almost-Linear Time

Xiaoyu He
Ray Li

The Longest Common Subsequence (LCS) is a fundamental string similarity measure, and computing the LCS of two strings is a classic algorithms question. A textbook dynamic programming algorithm gives an exact algorithm in quadratic time, and this is essentially best possible under plausible fine-grained complexity assumptions, so a natural problem is to find faster approximation algorithms. When the inputs are two binary strings, there is a simple 1/2-approximation in linear time: compute the longest common all-0s or all-1s subsequence. It has been open whether a better approximation is possible even in truly subquadratic time. Rubinstein and Song showed that the answer is yes under the assumption that the two input strings have equal lengths. We settle the question, generalizing their result to unequal length strings, proving that, for any ε>0, there exists δ>0 and a (1/2+δ)-approximation algorithm for binary LCS that runs in n 1+ε time. As a consequence of our result and a result of Akmal and Vassilevska-Williams, for any ε>0, there exists a (1/ q +δ)-approximation for LCS over q -ary strings in n 1+ε time.

Details

FOCS Conference 2021 Conference Paper

The zero-rate threshold for adversarial bit-deletions is less than 1/2

Venkatesan Guruswami
Xiaoyu He
Ray Li

We prove that there exists an absolute constant 6 > 0 such any binary code $C$ ⊂ {0, 1} N tolerating (1/2 - δ) $N$ adversarial deletions must satisfy| C| ≤ 2 polylog $N$ and thus have rate asymptotically approaching 0. This is the first constant fraction improvement over the trivial bound that codes tolerating $N$ /2 adversarial deletions must have rate going to 0 asymptotically. Equivalently, we show that there exists absolute constants $A$ and 6 > 0 such that any set $C$ ⊂ {0, 1} of 2 log A N binary strings must contain two strings $c$ and c’ whose longest common subsequence has length at least (1/2 + δ) N. As an immediate corollary, we show that q-ary codes tolerating a fraction 1 - (1 + 2δ) / $q$ of adversarial deletions must also have rate approaching 0. Our techniques include string regularity arguments and a structural lemma that classifies binary strings by their oscillation patterns. Leveraging these tools, we find in any large code two strings with similar oscillation patterns, which is exploited to find a long common subsequence.

Details

AIJ Journal 2019 Journal Article

A set of new multi- and many-objective test problems for continuous optimization and a comprehensive experimental evaluation

Yuren Zhou
Xiaoyu He
Yi Xiang
Shaowei Cai

Multi- and many-objective optimization problems have wide applications in the real world, and they have received growing attention from the evolutionary computation community. To promote the algorithm development in this area, numerous studies have been devoted to designing multi- and many-objective test problems. Most of these studies focus on handling complicated Pareto fronts (PFs), and the impact of the Pareto sets (PSs) has not been well-studied, although complicated PSs are prevalent in the real world. This paper presents a set of scalable test problems according to a new principle, which considers the geometrical properties of both PF and PS. A position function with a spherical form is proposed to introduce non-linear variable dependences to the PS, so as to simulate the variable dependencies in the real-world problems. According to the proposed principle, the first m (i. e. , the number of objectives) decision variables are used to form the surface of a unit hypersphere, while the rest variables are designed to optimize a certain distance function. A set of test problems are generated by the proposed principle, which are then used to investigate six representative algorithms. The experimental results indicate that the proposed test problems pose considerable difficulties to existing algorithms, calling for the need for designing new algorithms capable of handling complicated PF and PS simultaneously.

Details DOI

AAAI Conference 2019 Conference Paper

Running Time Analysis of MOEA/D with Crossover on Discrete Optimization Problem

Zhengxin Huang
Yuren Zhou
Zefeng Chen
Xiaoyu He

Decomposition-based multiobjective evolutionary algorithms (MOEAs) are a class of popular methods for solving multiobjective optimization problems (MOPs), and have been widely studied in numerical experiments and successfully applied in practice. However, we know little about these algorithms from the theoretical aspect. In this paper, we present a running time analysis of a simple MOEA with crossover based on the MOEA/D framework (MOEA/D-C) on four discrete optimization problems. Our rigorous theoretical analysis shows that the MOEA/D-C can obtain a set of Pareto optimal solutions to cover the Pareto front of these problems in expected running time apparently lower than the one without crossover. Moreover, the MOEA/D-C only needs to decompose an MOP into a few scalar optimization subproblems according to several simple weight vectors. This result suggests that the use of crossover in decomposition-based MOEA can simplify the setting of weight vector for different problems and make the algorithm more efficient. This study theoretically explains why some decomposition-based MOEAs work well in computational experiments and provides insights in design of MOEAs for MOPs in future research.

PDF Details