Parameterization¶
A parameter file is mandatory for the Machine Learning Trainer.
The file extension is ‘.prm’. All parameters must be given, even if they are not used. All parameters follow common tag-value notation. Rudimentary checks are performed by the software components using this file.
The ++PARAM_TRAIN_START++ and ++PARAM_TRAIN_END++ keywords enclose the parameter file.
The following parameter descriptions are a print-out of force-parameter, which can generate an empty parameter file skeleton.
Input
File that is holding the features for training (and probably validation). The file needs to be a table with features in columns, and samples in rows. Column delimiter is whitespace. The same number of features must be given for each sample. Do not include a header. The samples need to match the response file.
Type: full file pathFILE_FEATURES = NULLFile that is holding the response for training (class labels or numeric values). The file needs to be a table with one column, and samples in rows. Do not include a header. The samples need to match the feature file.
Type: full file pathFILE_RESPONSE = NULL
Output
File for storing the Machine Learning model in xml format. This file will be overwritten if it exists.
Type: full file pathFILE_MODEL = NULLFile for storing the logfile. This file will be overwritten if it exists.
Type: full file pathFILE_LOG = NULL
Training
Response variable for training the model. This number refers to the column of the response file, in which the desired variable is stored (FILE_RESPONSE).
Type: Integer. Valid range: [1,NUMBER_OF_VARIABLES]RESPONSE_VARIABLE = 1This parameter specifies how many samples (in %) should be used for training the model. The other samples are left out, and used to validate the model.
Type: Float. Valid range: ]0,100]PERCENT_TRAIN = 70This parameter specifies whether the samples should be randomly drawn (TRUE) or if the first n samples (FALSE) should be used for training.
Type: Logical. Valid values: {TRUE,FALSE}RANDOM_SPLIT = TRUEMachine learning method. Currently implemented are Random Forest and Support Vector Machines, both in regression and classification flavors.
Type: Character. Valid values: {SVR,SVC,RFR,RFC}ML_METHOD = RFCClass weights. This parameter only applies for the classification flavor. This parameter lets you define à priori class weights, which can be useful if the training data are inbalanced. This parameter can be set to a number of different values. EQUALIZED gives the same weight to all classes (default). PROPORTIONAL gives a weight proportional to the class frequency. ANTIPROPORTIONAL gives a weight, which is inversely proportional to the class frequency. Alternatively, you can use custom weights, i.e. a vector of weights for each class in your response file. The weights must sum to one, and must be given in ascending order.
Type: Character / Float list. Valid values: {EQUALIZED,PROPORTIONAL,ANTIPROPORTIONAL} or ]0,1[FEATURE_WEIGHTS = EQUALIZED
Random Forest parameters
This block only applies if method is Random Forest
Maximum number of trees in the forest. If RF_OOB_ACCURACY is 0, all trees will be grown. If RF_OOB_ACCURACY is set, the algorithm won’t grow additional trees if the accuracy is already met. If set to 0, additional trees are grown until RF_OOB_ACCURACY is met; note that this might never happen. RF_NTREE and RF_OOB_ACCURACY cannot both be 0.
Type: Integer. Valid range: [0,…RF_NTREE = 500Required accuracy of the ensemble, measured as OOB error. See also RF_NTREE.
Type: Float. Valid range: [0,…RF_OOB_ACCURACY = 0The number of randomly selected features at each tree node, which are used to find the best split. If set to 0, the square root of the number of all feature is used for the classification flavor; a third of the number of all feature is used for the regression flavor.
Type: Integer. Valid range: [0,…RF_NFEATURE = 0This parameter indicates whether the variable importance should be computed.
Type: Logical. Valid values: {TRUE,FALSE}RF_FEATURE_IMPORTANCE = TRUEIf the number of samples in a node is less than this parameter then the node will not be split. If set to 0, it defaults to 1 for the classification flavor; 5 for the regression flavor.
Type: Integer. Valid range: [0,…RF_DT_MINSAMPLE = 0The maximum possible depth of the tree. That is the training algorithms attempts to split a node while its depth is less than RF_DT_MAXDEPTH. The root node has zero depth. If set to 0, the maximum possible depth is used.
Type: Integer. Valid range: [0,…RF_DT_MAXDEPTH = 0Termination criteria for regression trees. If all absolute differences between an estimated value in a node and values of train samples in this node are less than this parameter then the node will not be split further.
Type: Float. Valid range: [0.01,…RF_DT_REG_ACCURACY = 0.01
Support Vector Machine parameters
This block only applies if method is Support Vector Machine
Maximum number of iterations for the iterative SVM training procedure which solves a partial case of constrained quadratic optimization problem. If SVM_ACCURACY is 0, all iterations will be used. If SVM_ACCURACY is set, the algorithm will stop if the accuracy is already met. If set to 0, additional iterations are computed until SVM_ACCURACY is met; note that this might never happen. SVM_MAXITER and SVM_ACCURACY cannot both be 0.
Type: Integer. Valid range: [0,…SVM_MAXITER = 1000000Required accuracy of the optimization. See also SVM_MAXITER.
Type: Float. Valid range: [0,…SVM_ACCURACY = 0.001Cross-validation parameter. The training set is divided into kFold subsets. One subset is used to test the model, the others form the train set. So, the SVM algorithm is executed kFold times.
Type: Float. Valid range: [1,…SVM_KFOLD = 10Parameter ϵ of a SVM optimization problem.
Type: Float. Valid range: [0,…SVM_P = 0Parameter C of a SVM optimization problem. This parameter expects three values which are used to perform a grid search, i.e. minimum value, maximum value, logarithmic step.
Type: Float list. Valid range: [0,…SVM_C_GRID = 0.001 10000 1Parameter γ of a kernel function.
This parameter expects three values which are used to perform a grid search, i.e. minimum value, maximum value, logarithmic step.
Type: Float list. Valid range: [0,…SVM_GAMMA_GRID = 0.000010 10000 10