Research map / Learning theory and algorithms / Clustering
Pursuit of the Cluster Structure of Network Lasso: Recovery Condition and Non-convex Extension
Shotaro Yagishita, Jun-ya Gotoh
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use network trimmed lasso instead of CFLA(S) | clustering of linear-model clients · 100 clients, N=4, nu=100 | max adjusted Rand index: 0.3913 -> 0.9289 | Improved |
| use network trimmed lasso instead of CFLA(G) | clustering of linear-model clients · 100 clients, N=8, nu=100 | max adjusted Rand index: 0.5568 -> 0.8189 | Improved |
| use NTL instead of CFLA(S) with abundant data | clustering of linear-model clients · 100 clients, N=8, nu=1000 | max adjusted Rand index: 0.8676 -> 0.9698 | Improved |
| use NTL instead of CFLA(G) with scarce data | clustering of linear-model clients · 100 clients, N=4, nu=10 | max adjusted Rand index: 0.4682 -> 0.2392 | Hurt |
| use NTL instead of CFLA(S) with abundant data | clustering of linear-model clients · 100 clients, N=4, nu=1000 | max adjusted Rand index: 0.9488 -> 0.9773 | Improved |
| use NTL instead of uniform-weight convex clustering | clustering · iris | max adjusted Rand index: 0.0015 -> 0.5778 | Improved |
| use NTL instead of weighted convex clustering w2 | clustering · digits (n = 500) | max adjusted Rand index: 0.5346 -> 0.3967 | Hurt |
| use NTL instead of weighted convex clustering w2 | clustering · wine | max adjusted Rand index: 0.7994 -> 0.8260 | Improved |
| sparser prior weights w2 instead of w1 (ablation) | convex clustering · digits (n = 100) | max adjusted Rand index: 0.4577 -> 0.5101 | Improved |
| use NTL instead of network lasso with ample prior weights | piecewise constant fitting · 1000 noisy signals | level estimation accuracy | Improved |
| use NTL instead of convex clustering w2 (computation) | ADMM runtime · digits (n = 500) | CPU seconds per ADMM run: 59.0054 -> 81.8792 | Hurt |
| initialize NTL from the network lasso midpoint solution (ablation) | NTL · two-cluster regression | cluster recovery | Improved |
Closest papers by meaning
- Simultaneous Clustering and Model Selection: Algorithm, Theory and Applications (TPAMI 2017 · similarity 0.87)
- Convex Clustering: Model, Theoretical Guarantee and Efficient Algorithm (JMLR 2021 · similarity 0.87)
- Graphical Lasso and Thresholding: Equivalence and Closed-form Solutions (JMLR 2019 · similarity 0.87)
- Variance estimation in graphs with the fused lasso (JMLR 2024 · similarity 0.86)
- LassoNet: A Neural Network with Feature Sparsity (JMLR 2021 · similarity 0.86)
- Statistically Optimal $K$-means Clustering via Nonnegative Low-rank Semidefinite Programming (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
