Beyond using multimodal features of items in addition to user-item interactions, researchers have additionally utilized Contrastive Learning (CL) in recent multimedia recommender systems to highly alleviate the data sparsity problem. CL-based methods generate at least two embeddings (i.e., views) for each instance and enrich the information of each instance from various perspectives via the views, thereby alleviating the data sparsity problem. Therefore, CL-based methods have focused on generating views that effectively represent the characteristics of each instance for their downstream tasks. Similarly, CL-based multimedia recommender systems have made efforts to effectively generate their user/item views by leveraging items' multimodal features. However, we point out the following two limitations that they have overlooked in generating their views: (1) they either have not attempted to identify the influence of each modality feature of an item on user-item interactions, or have identified it by randomly masking or dropping user-item interactions, and (2) they have not attempted to identify non-interactions likely to result in interactions in the future. To overcome these limitations, we propose a novel multimedia recommendation framework, named STARLINE, utilizing contraSTive leARning with modaLIty-aware graph refiNEment. Extensive experiments on five real-world datasets validate the effectiveness and validity of STARLINE, especially showing consistently higher accuracy by up to 13.24% compared to the best competitor.

This paper proposes transfer learning with spatiotemporal feature analysis using public facial images to build an automatic detection of facial paralysis caused by acute stroke. The overall process includes 1) facial detection and alignment network to extract major regions from the face, 2) transfer learning with feature extraction networks and gated recurrent unit, and 3) a classifier evaluating facial paralysis. We leveraged the Korean facial image data (K-FACE) from the public AI Hub to compensate for the insufficient data representing acute stroke symptoms. The experiment analyzed the effect of transfer learning and time series analysis using a gated recurrent unit with the deep learning models based on MobileNetV2, VGG16, and DenseNet121. Utilizing a facial big data system, transfer learning with spatiotemporal features showed a prominent performance with an accuracy of 0.925 and AUC of 0.924, which indicates the feasibility of real-time detection of stroke in daily living.