Abstract
Partitioning directed acyclic graphs is central to scheduling, pipelining, and memory-hierarchy optimization. We break the symmetry in classical spectral bi-partitioning in order to incentivize the alignment of directed graph cuts and find partitions whose cut edges mostly point in one direction. Our method employs a modified spectral quadratic objective function that promotes cut edges with consistent directionality. We further introduce an algorithm that rectifies misaligned edges to produce acyclic bi-partitions with low balanced cut metrics, and extend it to the retrieval of topological orders. We demonstrate the effectiveness of the proposed algorithms in comparative experiments on real-world cases that focus on acyclic density-based bi-partitioning.
| Original language | English |
|---|---|
| Title of host publication | Proceedings of the 38th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 26 |
| Publisher | Association for Computing Machinery |
| Pages | 482-485 |
| Number of pages | 4 |
| ISBN (Electronic) | 9798400727610 |
| DOIs | |
| State | Published - 8 Jul 2026 |
| Externally published | Yes |
| Event | 38th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 2026 - London, United Kingdom Duration: 6 Jul 2026 → 10 Jul 2026 |
Publication series
| Name | Annual ACM Symposium on Parallelism in Algorithms and Architectures |
|---|---|
| ISSN (Print) | 1548-6109 |
Conference
| Conference | 38th ACM Symposium on Parallelism in Algorithms and Architectures, SPAA 2026 |
|---|---|
| Country/Territory | United Kingdom |
| City | London |
| Period | 6/07/26 → 10/07/26 |
Bibliographical note
Publisher Copyright:© 2026 Copyright held by the owner/author(s).
Keywords
- (nearly) acyclic partitioning
- conductance
- directed acyclic graph (DAG) partitioning
- spectral partitioning
Fingerprint
Dive into the research topics of 'Brief Announcement: Direction-Incentivized Spectral Partitioning for Acyclic Graphs'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver