Sign Rank Limitations for Inner Product Graph Decoders
Abstract
Inner product-based decoders are among the most influential frameworks used to extract meaningful data from latent embeddings. However, such decoders have shown limitations in representation capacity in numerous works within the literature, which have been particularly notable in graph reconstruction problems. In this paper, we provide the first theoretical elucidation of this pervasive phenomenon in graph data, and suggest straightforward modifications to circumvent this issue without deviating from the inner product framework.
Cite
Text
Lee et al. "Sign Rank Limitations for Inner Product Graph Decoders." International Conference on Machine Learning, 2024.Markdown
[Lee et al. "Sign Rank Limitations for Inner Product Graph Decoders." International Conference on Machine Learning, 2024.](https://mlanthology.org/icml/2024/lee2024icml-sign/)BibTeX
@inproceedings{lee2024icml-sign,
title = {{Sign Rank Limitations for Inner Product Graph Decoders}},
author = {Lee, Su Hyeong and Zhang, Qingqi and Kondor, Risi},
booktitle = {International Conference on Machine Learning},
year = {2024},
pages = {27118-27136},
volume = {235},
url = {https://mlanthology.org/icml/2024/lee2024icml-sign/}
}