תקציר
We provide an analysis and interpretation of total variation (TV) minimization for semi-supervised learning from partially-labeled network-structured data. Our approach exploits an intrinsic duality between TV minimization and network flow problems. In particular, we use Fenchel duality to establish a precise equivalence of TV minimization and a minimum cost flow problem. This provides a link between modern convex optimization methods for non-smooth Lasso-Type problems and maximum flow algorithms. We show how a primal-dual method for TV minimization can be interpreted as distributed network optimization. Moreover, we derive a condition on the network structure and available label information that ensures that TV minimization accurately learns (approximately) piece-wise constant graph signals. This condition depends on the existence of sufficiently large network flows between labeled data points. We verify our analysis in numerical experiments.
שפה מקורית | אנגלית |
---|---|
מספר המאמר | 8902040 |
עמודים (מ-עד) | 6256-6269 |
מספר עמודים | 14 |
כתב עת | IEEE Transactions on Signal Processing |
כרך | 67 |
מספר גיליון | 24 |
מזהי עצם דיגיטלי (DOIs) | |
סטטוס פרסום | פורסם - 15 דצמ׳ 2019 |
פורסם באופן חיצוני | כן |