Explaining DNN using CBR as explanation engine.

Install packages in requirements.txt. The easiest is to use pip.

A guide on how the files should be structured.

CBR
 |__ libs
        |__ mycbr-sdk
                |__pom.xml
                |__src
                |__target
                    |__myCBR-x.x-SNAPSHOT.jar
        |__ mycbr-rest
                |__pom.xml
                |__src
                |__target
                    |__mycbr-rest-x.x-SNAPSHOT.jar
                |__lib/no/ntnu/mycbr/mycbr-sdk/

Download mycbr-rest and mycbr-sdk and structure them in folders as instructed above. Then, assuming you're in CBR/libs/mycbr-rest and have maven installed, run:

# Build sdk
cd ../mycbr-sdk
mvn clean install
# Build rest
cd ../mycbr-rest
mvn install:install-file -Dfile=../mycbr-sdk/target/myCBR-3.3-SNAPSHOT.jar -DpomFile=../mycbr-sdk/pom.xml -DlocalRepositoryPath=lib/no/ntnu/mycbr/mycbr-sdk/
mvn clean install

Not required, but helpful to view the project in a GUI. Can be found here: http://mycbr-project.org/download.html

Run these command to reproduce results in thesis. Note that these operation start up a myCBR server instance every time, so expect this to take some time.

Populate case-base with cases. Default is set to 5.

python main.py fill_final

Retrieve most similar case from the case-base.

python main.py retrieve

Retaining a test-case can be done by setting storage in the retrieve step to true. Default is false.

To run server standalone (still in CBR/libs/mycbr-rest):

# Operations are only persistent in memory
java -DMYCBR.PROJECT.FILE=/path/to/project.prj -jar ./target/mycbr-rest-1.0-SNAPSHOT.jar
# With the save option, all operations are saved to the .prj file
java -DMYCBR.PROJECT.FILE=/path/to/project.prj -Dsave=true -jar ./target/mycbr-rest-1.0-SNAPSHOT.jar