We use the ConceptBase system for analysing larger networks, e.g. enterprise models. Many queries rely on the transitive closure of some relations. If the graph has many loops (i.e. bi.directional transitive relations), then the DATALOG engine of ConceptBase gets to its knees.
Are there algorithms or heuristics known that deal specifically with this case? The DATALOG engine of ConceptBase does use tabling, i.e. it computes one fact only once. Still, graphs with many links pose a challenge.
Any hint is welcome!
Manfred
Hi.,
Datalog provides a way of incorporating recursion into a query language. ... our task can be viewed as solving the incremental transitive closure computation problem for Chain Datalog programs and generalizations allow special optimization techniques. . tational engine is structural recursion on sets.
Transitive closure and recursive Datalog implemented on clusters .... the right measure of algorithm efficiency for many applications, including ... Limitations of such techniques were also noticed Cos99, ACY95, .parallel Datalog engines on complex and large-scale Datalog evaluation, by a 4-6X margin.
Refer this link:
https://pdfs.semanticscholar.org/df1a/7d547359e89d8bb7d7bf9fbc084e8d5a6a4d.pdf
Since ConceptBase has its own Datalog-neg engine (based on tabling), the background of my question is to improve the performance for large graphs. My (naive) idea is that connected components of a graph could be precomputed and then be collapsed to a single virtual node when computing the transitive closure: each outside connection of the virtual node applies to all members when using the transitive closure of a relation. The problem is that the computation of the connected component itself is costly.
- Badges
- Science topic
- Similar topics
- Engineering
- Control Theory
- Control Systems
- Control Systems Engineering