P
Pong Eksombatchai
Publications - 5
Citations - 3181
Pong Eksombatchai is an academic researcher. The author has contributed to research in topics: Deep learning & Recommender system. The author has an hindex of 4, co-authored 4 publications receiving 1499 citations.
Papers
More filters
Proceedings ArticleDOI
Graph Convolutional Neural Networks for Web-Scale Recommender Systems
TL;DR: A novel method based on highly efficient random walks to structure the convolutions and a novel training strategy that relies on harder-and-harder training examples to improve robustness and convergence of the model are developed.
Proceedings ArticleDOI
Graph Convolutional Neural Networks for Web-Scale Recommender Systems.
TL;DR: In this paper, a data-efficient graph convolutional network (GCN) algorithm PinSage is proposed to generate embeddings of nodes that incorporate both graph structure as well as node feature information.
Proceedings ArticleDOI
Hierarchical Temporal Convolutional Networks for Dynamic Recommender Systems
TL;DR: HierTCN as discussed by the authors is a hierarchical deep learning architecture that makes dynamic recommendations based on users' sequential multi-session interactions with items, which is designed for web-scale systems with billions of items and hundreds of millions of users.
Posted Content
Hierarchical Temporal Convolutional Networks for Dynamic Recommender Systems
TL;DR: HierTCN, a hierarchical deep learning architecture that makes dynamic recommendations based on users' sequential multi-session interactions with items, is proposed and shown to be 2.5x faster than RNN-based models and uses 90% less data memory compared to TCN- based models.
Journal ArticleDOI
TransAct: Transformer-based Realtime User Action Model for Recommendation at Pinterest
Xue Xia,Pong Eksombatchai,Nikil Pancha,Dhruvil Badani,Po-Wei Wang,Nazanin Farahpour,Zhiyuan Zhang,Andrew Zhai +7 more
TL;DR: TransAct as mentioned in this paper is a sequential model that extracts users' short-term preferences from their real-time activities, which combines end-to-end sequential modeling via TransAct with batch-generated user embeddings.