Distributed Regular Path Query Matching and Optimization for Graph Database based on Spark
Bibliographic record
Abstract
We live in a world of connections where everything shares relationships like follow/subscribe in Social Network or protein interactions in Biology Network. A graph database embraces relationships and supports low-level join as its nature. Regular Path queries (RPQs) are queries run against graph database, which are written in the form of regular expressions based on edge labels and with strong flexibility and expressiveness. Unlike some graph databases where actual data stored and queried using standard relational mechanisms, in this thesis we investigate three distributed algorithms by storing graphs with NoSQL data model and evaluating RPQs with Apache Spark. The three algorithms are cascaded 2-way join, multi-way join and Dan Suciu’s Algorithm. The performance of them regarding to running time and network communication volume are compared, and main bottlenecks are identified. Dan Suciu’s algorithm shuffles the least data during evaluation, meanwhile the performance is heavily influenced by the ways of partitioning the graphs. In theory we found that the size of GAG (Global Accessible Graph) collected to driverside, which affects communication volume and computation scale on driver-side, is related to the number of input-nodes in distributed graph. So in this thesis project we also try to optimize the execution of Dan Suciu’s algorithm with various partition strategies such as METIS or JabeJa. Based on JabeJa, which tries to minimize the number of cross-edges, we propose a distributed algorithm JabeJa* to minimize the number of input-nodes in graph. In the best cases, those strategies can reduce the communication volume to 30%, driver-side compuation time to 30% and overall running time to 50%.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".