20170305-NOOR

Experiment design

Application of repair algorithms every 400 games (4 agents; 10000 iterations)

Hypotheses:

• Using repair systems does not improve F-measure whatever the level achieved by agents.

Experimental setting: Repetition of 20170227-NOOR with corrected software (clever was ineffective)

Experiment

Experimenter: Jérôme Euzenat (INRIA)

Date: 20170304-20170305

Lazy lavender hash: 41190f20796f7dfa6b221cdaf783580ed149f2d2

Parameters: params.sh

Command line (script.sh):

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=refadd -DrepeatRepairs=400 -Drepairers=logmap,alcomo

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=refadd -DrepeatRepairs=400 -Drepairers=logmap,alcomo -DexpandAlignments=clever -DnonRedundancy

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=refadd -DrepeatRepairs=400 -Drepairers=logmap,alcomo -DimmediateRatio=80

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=refadd -DrepeatRepairs=400 -Drepairers=logmap,alcomo -DexpandAlignments=clever -DnonRedundancy -DimmediateRatio=80

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

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

20170305-NOOR-4-10000-refadd-clever-nr-im80-rr.tsv 20170305-NOOR-4-10000-refadd-clever-nr-im80-rr.txt 20170305-NOOR-4-10000-refadd-clever-nr-rr.tsv 20170305-NOOR-4-10000-refadd-clever-nr-rr.txt 20170305-NOOR-4-10000-refadd-im80-rr.tsv 20170305-NOOR-4-10000-refadd-im80-rr.txt 20170305-NOOR-4-10000-refadd-rr.tsv 20170305-NOOR-4-10000-refadd-rr.txt

Result exploration

No expansion, no immediate consumption probability:

Memory expansion, no immediate consumption probability:

No expansion, immediate consumption probability 80%:

Memory expansion, immediate consumption probability 80%:

(drawn from data scaled every 400 measures)

Analysis

Key points: The results show that:

• as long as the agents have incoherent alignments, logical repair systems make them coherent at the expense of a loss in F-measure.
• as soon as the agents have coherent alignments, logical repair systems do nothing.

Further experiments:

• none

This file can be retrieved from URL https://sake.re/20170305-NOOR

It is possible to check out the repository by cloning https://felapton.inrialpes.fr/cakes/20170305-NOOR.git

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