Research map / Architectures and efficiency / Pruning and efficient tuning
Value Compression of Pattern Databases
Nathan Sturtevant, Ariel Felner, Malte Helmert
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 |
|---|---|---|---|
| compress PDB 2x with VC (4 bits) instead of EC 2x | IDA* with max of three 8-token PDBs · (18,4)-TopSpin | nodes expanded: 7.11M -> 3.88M | Improved |
| compress PDB 4x with VC (2 bits) instead of EC 4x | IDA* with BPMX · (18,4)-TopSpin | nodes expanded: 13.75M -> 10.39M | Improved |
| compress 2x with VC to 4 bits versus no compression | IDA* with BPMX · (18,4)-TopSpin | nodes expanded: 3.88M -> 3.88M | No clear effect |
| compress further to 3 bits (2.66x) with VC | IDA* with BPMX · (18,4)-TopSpin | nodes expanded: 3.88M -> 4.03M | No clear effect |
| compress to 2 bits (4x) with VC rather than 4 bits | IDA* with BPMX · (18,4)-TopSpin | nodes expanded: 3.88M -> 10.39M | Hurt |
| remove BPMX (ablation) | IDA* with VC 2-bit · (18,4)-TopSpin | nodes expanded | Hurt |
| train VC on static distribution (4s) versus EC 2x | IDA* · (18,2)-TopSpin | nodes expanded: 416,014 -> 3,645,502 | Hurt |
| train VC on dynamic distribution (4d) instead of static (4s) (ablation) | IDA* with VC 4 bits · (18,2)-TopSpin | nodes expanded: 3,645,502 -> 684,846 | Improved |
| manually tuned range selection (4c) versus EC 2x | IDA* with VC 4 bits · (18,2)-TopSpin | nodes expanded: 416,014 -> 394,603 | Improved |
| compress 2x with VC instead of EC | MM · (18,4)-TopSpin | nodes expanded: 3.55M -> 2.07M | Improved |
| compress 4x with VC (2 bits) instead of EC 4x | MM · (18,4)-TopSpin | nodes expanded: 5.42M -> 4.63M | Improved |
| combine EC 2x and VC 2x instead of VC 4x alone at 0.25 memory (ablation) | MM · (18,4)-TopSpin | nodes expanded: 4.63M -> 3.55M | Improved |
Closest papers by meaning
- Random Cycle Coding: Lossless Compression of Cluster Assignments via Bits-Back Coding (NeurIPS 2024 · similarity 0.83)
- Entropy Coding of Unordered Data Structures (ICLR 2024 · similarity 0.82)
- The Geometry of Linear Program Compression: An Exact Characterization and Learning Algorithm (NeurIPS 2026 · similarity 0.80)
- MeLLoC: Lossless Compression with High-order Mechanism Learning (NeurIPS 2024 · similarity 0.80)
- Practical Shuffle Coding (NeurIPS 2024 · similarity 0.80)
- Anchor Search: A Unified Framework for Suboptimal Bidirectional Search (AAAI 2025 · similarity 0.78)
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.
