-
Notifications
You must be signed in to change notification settings - Fork 0
3. Experiment 2 ‐ Recovering from degeneration
Jgocunha edited this page Oct 30, 2024
·
1 revision
-
git checkout recovering-from-degeneration
. - Open
dynamic-neural-field-degeneration
folder and runbuild.bat
. - The
recovering-from-degeneration.exe
will be located indynamic-neural-field-degeneration\dynamic-neural-field-degeneration\build\x64-release\Release\
.
You will only need to build the project once, afterwards if you want to change the experiment parameters, you just have to change the experiment_parameters.json
file. In this file, you will have the following options (hopefully, the names of the variables are pretty self-explanatory):
{
"experiment_parameters": {
"numberOfTrials": 100,
"startingExternalStimulus": 0,
"decisionTolerance": 2,
"isDataSavingOn": false,
"isVisualizationOn": true,
"isDebugModeOn": false
},
"degeneration_parameters": {
"#comment_type": "WEIGHTS_DEACTIVATE, WEIGHTS_RANDOMIZE, WEIGHTS_REDUCE, NEURONS_DEACTIVATE",
"experimentType": "WEIGHTS_DEACTIVATE",
"#comment_total_number": "720*280 = 201600 or 720 or 280",
"totalNumberOfElementsToDegenerate": 201600,
"#comment_fieldToDegenerate": "perceptual, output",
"fieldToDegenerate": "perceptual",
"#comment_numberOfElementsToDegeneratePerIteration": "1, 1, 1008",
"numberOfElementsToDegeneratePerIteration": 1008,
"#comment_initialPercentageOfDegeneration": "18, 88, 80",
"initialPercentageOfDegeneration": 80,
"targetPercentageOfDegeneration": 100,
"#comment_incrementOfDegenerationInPercentage": "0.1389, 0.278, 0.5",
"incrementOfDegenerationInPercentage": 0.5
},
"relearning_parameters": {
"#comment_type": "ALL_CASES, ONLY_DEGENERATED_CASES",
"relearningType": "ONLY_DEGENERATED_CASES",
"learningRate": 0.01,
"numberOfEpochs": 1,
"maxAmountOfDemonstrations": 200,
"#comment_update": "When deactivating neurons, updateAllWeights should be set to true",
"updateAllWeights": false
}
}
At dynamic-neural-field-degeneration\dynamic-neural-field-degeneration\data\results
you will find .txt
files with the amount of relearning cycles needed to recover lost behaviour as degeneration increases for each condition.
Additionally, you will find a analysis-files
folder with some .R scripts used to generate some plots (plots
folder) and some statistical analysis (analysis
folder).