20170215b-NOOR

Experiment design

Revision of networks of ontologies with immediate consumption probability and expansion with memory (4 agents; 10000 iterations)

Hypothesis:

• introducing the capability for agent to sometimes ignore the most specific applicable correspondences (with given immediate consumption probability) will improve precision by fighting shadowing.

Experimental setting: Repetition of part of 20170216-NOOR with consumption probability set to 80% (or repetition of 20170215a-NOOR with expansion-clever-nr)

Experiment

Experimenter: Jérôme Euzenat (INRIA)

Date: 20170215

Lazy lavender hash: dc342dca26a5a6feff895b2a1829b04e7dfaa5ae

Parameters: params.sh

Command line (script.sh):

for mod in delete replace refine add addjoin refadd
do
   java -Dlog.level=INFO -cp lib/lazylav/ll.jar:lib/slf4j/logback-classic-1.1.9.jar:lib/slf4j/logback-core-1.1.9.jar:. fr.inria.exmo.lazylavender.engine.Monitor -DnbAgents=4 -DnbIterations=10000 -DnbRuns=10 -DreportPrecRec -DrevisionModality=${mod} -DexpandAlignments=clever -DnonRedundancy -DimmediateRatio=80
done

Class used: NOOEnvironment, AlignmentAdjustingAgent, AlignmentRevisionExperiment, ActionLogger, AverageLogger, Monitor

Execution environment: Debian Linux virtual machine configured with four processors and 20GB of RAM running under a Dell PowerEdge T610 with 4*Intel Xeon Quad Core 1.9GHz E5-2420 processors, under Linux ProxMox 2 (Debian). - Java 1.8.0 HotSpot

Raw results

20170215b-NOOR-4-10000-add-clever-nr-im80.tsv 20170215b-NOOR-4-10000-add-clever-nr-im80.txt 20170215b-NOOR-4-10000-addjoin-clever-nr-im80.tsv 20170215b-NOOR-4-10000-addjoin-clever-nr-im80.txt 20170215b-NOOR-4-10000-delete-clever-nr-im80.tsv 20170215b-NOOR-4-10000-delete-clever-nr-im80.txt 20170215b-NOOR-4-10000-refadd-clever-nr-im80.tsv 20170215b-NOOR-4-10000-refadd-clever-nr-im80.txt 20170215b-NOOR-4-10000-refine-clever-nr-im80.tsv 20170215b-NOOR-4-10000-refine-clever-nr-im80.txt 20170215b-NOOR-4-10000-replace-clever-nr-im80.tsv 20170215b-NOOR-4-10000-replace-clever-nr-im80.txt

Result exploration

• In plain, average success rate (over 10 runs);
• In dashed, the semantic F-measure.

Comparison for F-measure with 20170216-NOOR:

• In plain, memory expansion operator semantic F-measure;
• In dashed, memory expansion+80% immediate consumption probability semantic F-measure.

(drawn from data scaled every 100 measures)

Analysis

Key points:

• The results converge in all cases to 100% precision and 0% incoherence.
• refadd has the best F-measure so far while the others have lower F-measure and recall than in the simple expansion with memory!
• The behaviour of addjoin as related to add is still unclear (see 20180530-NOOR)

Further experiments:

• sortout the add/addjoin discrepancy

This file can be retrieved from URL https://sake.re/20170215b-NOOR

It is possible to check out the repository by cloning https://felapton.inrialpes.fr/cakes/20170215b-NOOR.git

This experiment has been transferred from its initial location at https://gforge.inria.fr (not available any more)