Image-text matching task aims to learn the fine-grained correspondences between images and sentences. Existing methods use attention mechanism to learn the correspondences by attending to all fragments without considering the relationship between fragments and global semantics, which inevitably lead to semantic misalignment among irrelevant fragments. To this end, we propose a Bidirectional Correct Attention Network (BCAN), which leverages global similarities and local similarities to reassign the attention weight, to avoid such semantic misalignment. Specifically, we introduce a global correct unit to correct the attention focused on relevant fragments in irrelevant semantics. A local correct unit is used to correct the attention focused on irrelevant fragments in relevant semantics. Experiments on Flickr30K and MSCOCO datasets verify the effectiveness of our proposed BCAN by outperforming both previous attention-based methods and state-of-the-art methods. Code can be found at: https://github.com/liuyyy111/BCAN.

No abstract

No abstract

This paper researches the possibility of using locally weighted algorithm for intelligent modeling of a nonlinear system for vanadium extraction in metallurgical process and proposes some optimized methods by finding the optimized regression coefficients by gradient descent and kernel function bandwidth by weighted distance. But kernel matrix computation for high dimensional data source demands heavy computing power. To overcome the computational difficulties of kernel functions and shorten the computing time, the paper designs a distributed algorithm to compute the kernel function matrix of LWA. The paper then implements the algorithm on a cluster of computing workstations using MPI. This paper studies the possibility of LWA using distributed kernel computing for predictive modeling for vanadium extraction in metallurgical process. Finally, the practical data are used to study the speedups and accuracy of the algorithm. The experimental results show that optimized locally weighted algorithm using distributed kernel outperforms the traditional RBF, RFWR and LWPR methods when significant amounts of noise are added, and the computing time has been shortened. Index Terms -locally weighted algorithm; distributed kernel computing; gradient descent; weighted distance; intelligent model II. LOCALLY WEIGHTED ALGORITHM AND OPTIMIZED REGRESSION COEFFICIENTSLocally weighted algorithm is derived from standard linear regression. This algorithm fits a surface to "local" points using distance-weighted regression. The locally weighted algorithm 978-1-4244-2114-5/08/$25.00

No abstract