The increasing number of available Web Application Programming Interfaces (APIs) in various service sharing communities have enabled software developers to develop their interested multimedia mashups quickly and conveniently. In this situation, a multimedia mashup with complex functionalities could be achieved by composing a set of pre-selected Web APIs by software developers. However, due to the APIs diversity in terms of development organization, programming language, invocation interface, etc, it is often difficult to determine the compatibility between the APIs selected by multimedia mashup developers beforehand especially when the developers have little background knowledge of APIs, which significantly decreases the successful rate of subsequent multimedia mashup development. In response to this challenge, we propose a subgraph matching-based compatible API’s composition recommendation method, called SubMC_WACR. The advantage of SubMC_WACR is that it can directly search for the API’s subgraphs that not only meet the functional requirements of the multimedia mashup but also are compatible with each other, thus boosting the effectiveness of multimedia mashup development. Through extensive experiments on a real dataset crawled from the Web API sharing platform ProgrammableWeb.com, we have demonstrated that our proposed recommendation method achieves significant improvements in terms of recommendation precision and compatibility compared with other competitive API recommendation methods.

The rapid expansion of Web APIs presents developers with significant challenges in selecting optimal API compositions. To address this issue, keyword-based API composition recommendation techniques have been proposed. However, these methods often suffer from popularity bias due to the influence of historical datasets and recommendation models. This bias leads to the disproportionate recommendation of popular APIs over less popular ones, potentially causing the Matthew effect and impeding the balanced development of the API ecosystem. Although several studies have identified and attempted to mitigate popularity bias, they have largely relied on static analysis without accounting for the dynamic nature of API recommendations. In this paper, we introduce a dynamic simulation framework for API composition recommendations, designed to explore the evolution of popularity bias within recommendation results, and propose a debiasing method for dynamic recommendations by combining the enhanced API correlation graph with the Determinantal Point Process (DPP) method. Finally, extensive experiments on real datasets show that the algorithm effectively alleviates the popularity bias problem while guaranteeing high recommendation accuracy.

Sport plays a crucial role in society, influencing physical health, entertainment, and community engagement. As artificial intelligence advances, the ability to classify sport images accurately becomes increasingly crucial. Effective sport image classification enhances applications, such as performance analysis, athlete tracking, and fan engagement. Despite its significance, current methods face challenges due to limited labeled datasets and issues with feature misalignment. This paper introduces a novel Contrastive Language-Image Pre-training (CLIP) based framework specifically designed for sport image classification. By incorporating data augmentation techniques, the approach addresses data sparsity and enriches the diversity of image-text pairings, reducing the need for extensive manual annotation. Additionally, feature alignment strategies tackle text-image misalignment issues that affect classification accuracy. This approach fills a significant research gap and offers practical solutions to improve classification performance in sport image analysis. The results of extensive experiments validate the effectiveness of the framework, demonstrating its potential to advance sports analytics and contribute to more precise and scalable solutions in sport image classification.

Finding more specific subcategories within a larger category is the goal of fine-grained image classification (FGIC), and the key is to find local discriminative regions of visual features. Most existing methods use traditional convolutional operations to achieve fine-grained image classification. However, traditional convolution cannot extract multi-scale features of an image and existing methods are susceptible to interference from image background information. Therefore, to address the above problems, this paper proposes an FGIC model (Attention-PCNN) based on hybrid attention mechanism and pyramidal convolution. The model feeds the multi-scale features extracted by the pyramidal convolutional neural network into two branches capturing global and local information respectively. In particular, a hybrid attention mechanism is added to the branch capturing global information in order to reduce the interference of image background information and make the model pay more attention to the target region with fine-grained features. In addition, the mutual-channel loss (MC-LOSS) is introduced in the local information branch to capture fine-grained features. We evaluated the model on three publicly available datasets CUB-200-2011, Stanford Cars, FGVC-Aircraft, etc. Compared to the state-of-the-art methods, the results show that Attention-PCNN performs better.