A5027525032- Advanced Graph Neural Networks
- Recommender Systems and Techniques
- Domain Adaptation and Few-Shot Learning
- Explainable Artificial Intelligence (XAI)
- Text and Document Classification Technologies
- Software-Defined Networks and 5G
- Advanced Optical Network Technologies
- Power Line Communications and Noise
- Complex Network Analysis Techniques
- Bioinformatics and Genomic Networks
- Optical Network Technologies
- Graph Theory and Algorithms
- Consumer Market Behavior and Pricing
- Advanced Bandit Algorithms Research
- Epigenetics and DNA Methylation
- Adversarial Robustness in Machine Learning
- Machine Learning and Data Classification
- Machine Learning in Healthcare
- Advanced Wireless Communication Techniques
- Multimodal Machine Learning Applications
- PAPR reduction in OFDM
- Image and Video Quality Assessment
Samsung (United States)
2023
Samsung (South Korea)
2019-2022
Self-supervised learning (SSL) has been demonstrated to be effective in pre-training models that can generalized various downstream tasks. Graph Autoencoder (GAE), an increasingly popular SSL approach on graphs, widely explored learn node representations without ground-truth labels. However, recent studies show existing GAE methods could only perform well link prediction tasks, while their performance classification tasks is rather limited. This limitation casts doubt the generalizability...
Graph neural networks (GNNs) have received remarkable success in link prediction (GNNLP) tasks. Existing efforts first predefine the subgraph for whole dataset and then apply GNNs to encode edge representations by leveraging neighborhood structure induced fixed subgraph. The prominence of GNNLP methods significantly relies on adhoc Since node connectivity real-world graphs is complex, one shared limited all edges. Thus, choices subgraphs should be personalized different However, performing...
Graph neural networks (GNNs) integrate deep architectures and topological structure modeling in an effective way. However, the performance of existing GNNs would decrease significantly when they stack many layers, because over-smoothing issue. Node embeddings tend to converge similar vectors keep recursively aggregating representations neighbors. To enable GNNs, several methods have been explored recently. But are developed from either techniques convolutional or heuristic strategies. There...
Recommender systems play a fundamental role in web applications filtering massive information and matching user interests. While many efforts have been devoted to developing more effective models various scenarios, the exploration on explainability of recommender is running behind. Explanations could help improve experience discover system defects. In this paper, after formally introducing elements that are related model explainability, we propose novel explainable recommendation through...
Real-time bidding (RTB) that features perimpression-level real-time ad auctions has become a popular practice in today's digital advertising industry. In RTB, click-through rate (CTR) prediction is fundamental problem to ensure the success of an campaign and boost revenue. this paper, we present dynamic CTR model designed for Samsung demand-side platform (DSP). From our production data, identify two key technical challenges have not been fully addressed by existing solutions: nature RTB user...
Learning useful interactions between input features is crucial for tabular data modeling. Recent efforts start to explicitly model the feature with graph, where each treated as an individual node. However, existing graph construction methods either heuristically formulate a fixed feature-interaction based on specific domain knowledge, or simply apply attention function compute pairwise similarities sample. While may be sub-optimal downstream tasks, sample-wise time-consuming during training...
The recent breakthrough achieved by graph neural networks (GNNs) with few labeled data accelerates the pace of deploying GNNs on real-world applications. While several efforts have been made to scale training for large-scale graphs, still suffer from scalability challenge model inference, due dependency issue incurred message-passing mechanism, therefore hindering its deployment in resource-constrained An intuitive remedy is compressing cumbersome GNN into inference-friendly multi-layer...
Embedding is widely used in recommendation models to learn feature representations. However, the traditional embedding technique that assigns a fixed size all categorical features may be suboptimal due following reasons. In domain, majority of features' embeddings can trained with less capacity without impacting model performance, thereby storing equal length incur unnecessary memory usage. Existing work tries allocate customized sizes for each usually either simply scales feature's...
Recommender systems play a fundamental role in web applications filtering massive information and matching user interests. While many efforts have been devoted to developing more effective models various scenarios, the exploration on explainability of recommender is running behind. Explanations could help improve experience discover system defects. In this paper, after formally introducing elements that are related model explainability, we propose novel explainable recommendation through...