In this section
You can also access our individual websites (via the Members page) for further information about our research and lists of our publications.
@article{Qian:2022:10.1109/TNNLS.2020.3043196,
author = {Qian, Y and Expert, P and Rieu, T and Panzarasa, P and Barahona, M},
doi = {10.1109/TNNLS.2020.3043196},
journal = {IEEE Transactions on Neural Networks and Learning Systems},
pages = {1663--1672},
title = {Quantifying the alignment of graph and features in deep learning},
url = {http://dx.doi.org/10.1109/TNNLS.2020.3043196},
volume = {33},
year = {2022}
}
TY - JOUR
AB - We show that the classification performance of graph convolutional networks (GCNs) is related to the alignment between features, graph, and ground truth, which we quantify using a subspace alignment measure (SAM) corresponding to the Frobenius norm of the matrix of pairwise chordal distances between three subspaces associated with features, graph, and ground truth. The proposed measure is based on the principal angles between subspaces and has both spectral and geometrical interpretations. We showcase the relationship between the SAM and the classification performance through the study of limiting cases of GCNs and systematic randomizations of both features and graph structure applied to a constructive example and several examples of citation networks of different origins. The analysis also reveals the relative importance of the graph and features for classification purposes.
AU - Qian,Y
AU - Expert,P
AU - Rieu,T
AU - Panzarasa,P
AU - Barahona,M
DO - 10.1109/TNNLS.2020.3043196
EP - 1672
PY - 2022///
SN - 1045-9227
SP - 1663
TI - Quantifying the alignment of graph and features in deep learning
T2 - IEEE Transactions on Neural Networks and Learning Systems
UR - http://dx.doi.org/10.1109/TNNLS.2020.3043196
UR - http://hdl.handle.net/10044/1/85928
VL - 33
ER -