The integration of Large Language Models (LLMs) has led to substantial advancements in recommender systems (RS) by leveraging their vast knowledge and reasoning abilities. However, the semantic gap between the linguistic knowledge of LLMs and the collaborative patterns in RS hinders their effective fusion. This issue results in a fundamental limitation where models, despite achieving high prediction accuracy, are unable to provide coherent rationales justifying their recommendations. In this paper, we propose ARROW (Adaptive Reasoning for LLM-based RecommendatiOn With explainability), a novel framework that effectively elicits the intrinsic reasoning capabilities of LLMs to bridge this semantic gap. ARROW is carefully designed to guide the model in generating an explicit reasoning process for its recommendation decisions using chain-of-thought prompting. Furthermore, we introduce the Adaptive Reasoning Modulator, which quantifies the uncertainty of the reasoning process and adaptively adjusts its weight to maximize the model's reasoning efficacy. Our extensive experiments demonstrate that ARROW achieves significant performance improvements over strong baseline models while providing human-interpretable explanations. Our code is available at https://github.com/yunwooseong/ARROW.

Link prediction is an important learning task for graph-structured data, and has become increasingly popular due to its wide application areas. Graph Neural Network (GNN)-based approaches including Variational Graph Autoencoder (VGAE) have achieved promising performance on link prediction outperforming conventional models which use hand-crafted features. VGAE learns latent node representations and predicts links based on the similarities between nodes. While the inner product based decoder effectively utilizes the node representations for link prediction, it exhibits sub-optimal performance due to the intrinsic limitation of the inner product. We found that the the cosine similarity and norm simultaneously try to explain the link probability, which hinders the gradient flow during training. We also point out the message passing scheme is unexpectedly dominated by the nodes with large norm values. In this paper, we propose a stochastic VGAE-based method that can effectively decouple the norm and angle in the embeddings. Specifically, we relate the cosine similarity and norm to two fundamental principles in graph: homophily and node popularity respectively. Our learning scheme is based on a hard expectation maximization learning method; we infer which of the two has been exerted for link formation, and subsequently optimize based on this guess. Through extensive experiments on real-world datasets, we demonstrate our model outperforms the existing state-of-the-art methods on link prediction and achieves comparable performances on other downstream tasks such as node classification and clustering. Our code is at https://github.com/yoonsikcho/d-vgae.

Human Activity Recognition (HAR) with inertial sensors is one of the most active research fields. Various machine learning algorithms have been proposed in HAR for classifying human activities. However, these methods heavily rely on the quality of hand-crafted features, requiring extensive feature engineering. Recent deep learning approaches have tried to perform training in an end-to-end manner. We propose a new learning scheme based on Supervised Autoencoder with Self-Attention (SAE-SA). Our main idea is two-fold: (1) Through the dimensional reduction in supervised autoencoder, our model is robust to noisy input signals sensor data (2) We incorporate the self-attention mechanism in the classifier layer of supervised autoencoder, which focuses more to the relevant signals, and can learn the features without any knowledge of the aggregated signal data. We evaluate SAE-SA on benchmark datasets: WISDM v2.0, and PAMAP2. We achieve accuracy of 95.76%, and 97.60% respectively, which is the state-of-the-art results in HAR.

Text-to-video(T2V) retrieval aims to find relevant videos from text queries. The recently introduced Contrastive Language Image Pretraining (CLIP), a pretrained language-vision model trained on large-scale image and caption pairs, has been extensively studied in the literature for this task. Existing studies on T2V task have aimed to transfer the CLIP knowledge and focus on enhancing retrieval performance through fine-grained representation learning. While fine-grained contrast has achieved some remarkable results, less attention has been paid to coarse-grained contrasts. To this end, we propose a method called Graph Patch Spreading (GPS) to aggregate patches across frames at the coarse-grained level. We apply GPS to our proposed framework called Multi-Encoder Multi-Expert (MEME) framework. Our proposed scheme is general enough to be applied to any existing CLIP-based video-text retrieval models. We demonstrate the effectiveness of our method on existing models over the benchmark datasets MSR-VTT, MSVD, and LSMDC datasets. Our code can be found at https://github.com/kang7734/MEME__.

Motivated by the recent successes of deep generative models used for collaborative filtering, we propose a novel framework of VAE for collaborative filtering using multiple experts and stochastic expert selection, which allows the model to learn a richer and more complex latent representation of user preferences. In our method, individual experts are sampled stochastically at each user-item interaction which can effectively utilize the variability among multiple experts. While we propose this framework in the context of collaborative filtering, the proposed stochastic expert technique can be used to enhance VAEs in general beyond the application of collaborative filtering. Hence, this novel technique can be of independent interest. We comprehensively evaluate our proposed method, Stochastic-Expert Variational Autoencoder (SE-VAE) on numerical experiments on the real-world benchmark datasets from MovieLens and Netflix and show that it consistently outperforms the existing state-of-the-art methods across all metrics. Our proposed stochastic expert framework is generic and adaptable to any VAE architecture. The experimental results show that the adaptations to various architectures provided performance gains over the existing methods.

With the emergence of social networking services, researchers enjoy the increasing availability of large-scale heterogenous datasets capturing online user interactions and behaviors. Traditional analysis of techno-social systems data has focused mainly on describing either the dynamics of social interactions, or the attributes and behaviors of the users. However, overwhelming empirical evidence suggests that the two dimensions affect one another, and therefore they should be jointly modeled and analyzed in a multi-modal framework. The benefits of such an approach include the ability to build better predictive models, leveraging social network information as well as user behavioral signals. To this purpose, here we propose the Constrained Latent Space Model (CLSM), a generalized framework that combines Mixed Membership Stochastic Blockmodels (MMSB) and Latent Dirichlet Allocation (LDA) incorporating a constraint that forces the latent space to concurrently describe the multiple data modalities. We derive an efficient inference algorithm based on Variational Expectation Maximization that has a computational cost linear in the size of the network, thus making it feasible to analyze massive social datasets. We validate the proposed framework on two problems: prediction of social interactions from user attributes and behaviors, and behavior prediction exploiting network information. We perform experiments with a variety of multi-modal social systems, spanning location-based social networks (Gowalla), social media services (Instagram, Orkut), e-commerce and review sites (Amazon, Ciao), and finally citation networks (Cora). The results indicate significant improvement in prediction accuracy over state of the art methods, and demonstrate the flexibility of the proposed approach for addressing a variety of different learning problems commonly occurring with multi-modal social data.