Cross-modal retrieval refers to identifying semantically relevant data across different modalities. However, annotation errors or inherent ambiguity can cause semantic inconsistency in sample pairs, degrading retrieval performance. Prior efforts either relied heavily on the quality of explicitly dividing clean and noisy subsets, or solely leveraged carefully selected single anchor information, neglecting relationships among diverse neighbors. In this paper, we propose a novel Graph-based Label Propagation (GLP) framework that learns pseudo-labels via label propagation on a sparse graph, enabling self-correction of noisy labels. Specifically, each modality's instances are treated as nodes, connected via k-nearest neighbor (kNN) search to form a sparse graph. Pseudo-label vectors are generated for all nodes within one modality to capture the matching degree of inter-modal nodes. Through iterative label propagation, the stabilized pseudo-labels implicitly exploit both intra- and inter-modal relationships to derive a reliable matching degree. A dynamic queue further enhances graph quality by updating high-quality nodes. Experiments on Flickr30K, MSCOCO, and CC120K show that our method outperforms state-of-the-art approaches, especially under high noise. Code is available at https://github.com/njustkmg/MM25-GLP.

It is widely known that the error analysis for deep learning involves approximation, statistical, and optimization errors. However, it is challenging to combine them together due to overparameterization. In this paper, we address this gap by providing a comprehensive error analysis of the Deep Ritz Method (DRM). Specifically, we investigate a foundational question in the theoretical analysis of DRM under the overparameterized regime: given a target precision level, how can one determine the appropriate number of training samples, the key architectural parameters of the neural networks, the step size for the projected gradient descent optimization procedure, and the requisite number of iterations, such that the output of the gradient descent process closely approximates the true solution of the underlying partial differential equation to the specified precision?

Microaggressions are subtle, everyday comments or actions that communicate discrimination towards historically marginalized groups. While these can be small experiences, they can cause measurable harm. Researchers have investigated how to reduce the negative experiences that result. But there have only been limited investigations into how to support those efforts with technology. To address this, we conduct a series of fifteen participatory design workshops aimed at designing to reduce the occurrence of and negative impacts from microaggressions. The workshops included 47 participants drawn from communities frequently targeted by microaggressions, with groups formed around gender, race/ethnicity, and disability & accessibility. Our study findings identified four primary themes in designing effective interventions for microaggressions: proactive measures, reward and accountability frameworks, community support mechanisms, and long-term educational resources. Through axial coding, we observed that different groups respond uniquely to these intervention approaches, with distinct preferences for timing, intervention style, and accountability. We identify challenges participants face when intervening around microaggressions and suggest directions for future solutions.

Efforts to bring incarcerated and formerly incarcerated individuals into the field of computing stand to improve equitable access to both computing jobs, and consequently the benefits of our tools and innovations through the inclusion of more diverse perspectives. This report describes the design and execution of a college level introductory computing course conducted with 26 students currently incarcerated at a prison in the United States in Fall 2022. We discuss the ways that the prison environment and the student body differ from traditional college computing classes, and how this impacted the design and execution of the course. We found that despite significant environmental barriers to learning to program, such as not having access to a code interpreter, there were unique affordances of the student population, including maturity and community, that could be leveraged in the course design and policies. We conclude with many lessons learned for the purpose of improving future offerings of computing courses in prisons.

Previous research has been devoted to improving the experience of non-majors in introductory CS courses. In this study, we compare the experiences of non-majors in two different introductory CS courses, specifically with respect to fears about taking the course and change in confidence levels. CS0 is a computing course intentionally designed for non-majors, and CS1 is a more traditional introductory computing course. Both of these courses were composed primarily of non-majors and were taught by the same instructor. Survey data was collected from 124 students enrolled in CS0, and 502 students enrolled in CS1. Through qualitative analysis, we found that the fears of non-major students entering both of these introductory CS courses fell into one or more of nine distinct categories (e.g., Coding, Perceiving STEM as Difficult, Managing Workload). Additionally, using students' confidence levels at the beginning and end of the courses, we found that students in CS0 had a greater increase in confidence level than those in CS1. Finally, we explored connections between students' fears and how their confidence changed by the end of the course. We found that students across both courses with fears related to coding, lack of preparation, and being left behind had the highest average increase in confidence levels.

The facial pore feature is one of the crucial indicators for face recognition and skin evaluation. However, pores are tiny, which are difficult to detect and analysis based on the digital image. We proposed a new facial pore detection algorithm that combines the characteristics of skin pigment distribution and optimal scale, which can effectively eliminate the effect of complicated skin interferences from facial pore detection processing. First, considering the dissimilarity of skin pigment distribution on different pigment layers, we used SURF and SIFT algorithms to detect the skin features on different pigment layers and calculated the threshold by DBSCAN. Then, the Euclid distance was calculated to describe the positions similarity of the same detected points on different layers. Last, the optimal scales were set as thresholds to screen off the interferences of skin features except for pores. The experiment results confirm the improvement of the pore detection accuracy.