Minimal execution:

java -jar EccCL.jar -inpData PATH

performs training and classification with data in PATH.

java -jar EccCL.jar -inpGenI i -inpGenA a -inpGenL l

performs training and classification with randomly generated data with i instances, with each having a attributes and l labels.

For evaluation add -evalAllLabels or -evalPerLabel e.g.:

java -jar EccCL.jar -evalAllLabels -inpData PATH
java -jar EccCL.jar -evalPerLabel -inpGenI i -inpGenA a -inpGenL l

---
-buildOnly PATH
only training

-classOnly PATH
only classification
---
-inpData PATH
PATH to input data

-inpRep n
bootstrapping of input data

-inpNAL n
explicit statement that input has n attributes (only if .xml is missing)

-inpGenI n
generates n instances with random values. Has to be used with -inpGenA and -inpGenL

-inpGenA n
generates instances with n attributes. Has to be used with -inpGenI and -inpGenL

-inpGenL n
generates instances with n labels. Has to be used with -inpGenI and -inpGenA

-inpTrainMax n
use only n instances for training

-inpClassMax n
use only n instances for classification

-inpFillI n
fills data with random instances
---
-clLIB NAME
NAME of the native OpenCL-library.

-clNAT PFAD
native OpenCL-library can be found in PATH

-clWG n
use n work items per work group

-clNOFLOAT
use implementation with doubles

-clLOG
use Build-Log of OpenCL

-clDEV d
specifies CL Device Type (d = 0 for GPU, 1 for CPU)

-clMEM d
specifies CL Mem Type (d = 0 for Copy Host-Pointer, 1 for Use Host-Pointer).

-clKTIME
prints execution time of kernels

-clECCTIME
prints execution time of ECC
---
-eccExBuild d
specifies the modus for training (d = 0 for OpenCL/GPU, 1 for Single-threaded Java/CPU, 2 for Multi-threaded Java).

-eccExClass d
specifies the modus for classification (d = 0 for OpenCL/GPU, 1 for Single-threaded Java/CPU, 2 for Multi-threaded Java).

-eccML n
maximum depth of trees

-eccMA n
maximum number of nodes per attribute

-eccFS n
number of trees

-eccFSRun n
number of random forests per run

-eccES
number of classifier chains

-eccESRun n
number of chains trained per run

-eccESF s
proportion that is used for training of the chain. Between 0 (nothing) and 1 (complete input)

-eccFSF s
proportion that is used for training of the random forest. Between 0 (nothing) and 1 (complete input)

-eccInstClass n
number of instances to be classified per run

---
-benchREP n
Benchmark-Modus: repeat n times

-benchBalance s
Benchmark-Modus: proportion used for training

-benchThreads n
Benchmark-Modus: number of threads

-benchCSV PATH
Benchmark-Modus: saves benchmark results in PATH

-benchCSVMETA PATH
Benchmark-Modus: saves used parameters in PATH
---
-evalPerLabel
Evaluation per label

-evalAllLabels
evaluation with all labels