Representation Learning with Weighted Inner Product for Universal Approximation of General Similarities
Abstract
We propose weighted inner product similarity (WIPS) for neural network-based graph embedding. In addition to the parameters of neural networks, we optimize the weights of the inner product by allowing positive and negative values. Despite its simplicity, WIPS can approximate arbitrary general similarities including positive definite, conditionally positive definite, and indefinite kernels. WIPS is free from similarity model selection, since it can learn any similarity models such as cosine similarity, negative Poincaré distance and negative Wasserstein distance. Our experiments show that the proposed method can learn high-quality distributed representations of nodes from real datasets, leading to an accurate approximation of similarities as well as high performance in inductive tasks.
Cite
Text
Kim et al. "Representation Learning with Weighted Inner Product for Universal Approximation of General Similarities." International Joint Conference on Artificial Intelligence, 2019. doi:10.24963/IJCAI.2019/699Markdown
[Kim et al. "Representation Learning with Weighted Inner Product for Universal Approximation of General Similarities." International Joint Conference on Artificial Intelligence, 2019.](https://mlanthology.org/ijcai/2019/kim2019ijcai-representation/) doi:10.24963/IJCAI.2019/699BibTeX
@inproceedings{kim2019ijcai-representation,
title = {{Representation Learning with Weighted Inner Product for Universal Approximation of General Similarities}},
author = {Kim, Geewook and Okuno, Akifumi and Fukui, Kazuki and Shimodaira, Hidetoshi},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {2019},
pages = {5031-5038},
doi = {10.24963/IJCAI.2019/699},
url = {https://mlanthology.org/ijcai/2019/kim2019ijcai-representation/}
}