DylMath module

DylMath.MSE(sep: float, dist: str, ROC: list, rocEmpiric: list = None) → Tuple[float, float, float]

Returns the MSE of the given ROC.

If sep and dist are not None: the true ROC from sep and dist If rocEmpiric is not None: the MSE between the Empiric and ROC If sep and dist are None, the first value returned is always 0 The last value returned is always the AUC of the ROC

DylMath.auc(results: tuple, D0: list = None, D1: list = None) → float

Takes an ROC curve from genROC and returns the AUC.

If results is a prediction not an ROC curve, generates the ROC curve. If results is already an ROC curve, D0 and D1 are not required.

DylMath.avROC(rocs: list) → tuple

Averages ROC curves. Rocs parameter are ROC curves from genROC.

DylMath.calcNLayers(arr: list) → int

Returns the number of layers that would be needed to sort.

If arr is the a tuple or list, uses the length. If arr is already the length, uses that.

DylMath.genD0D1(d0d1: list, arr: list) → tuple

Generates filtered D0 and D1 vectors.

d0d1 is (D0, D1) together as a tuple/list.

DylMath.genROC(predicted: tuple, D1: list = None, D0: list = None) → list

Returns a list of collections of x,y coordinates in order of the threshold.

DylMath.genSep(dist: str, auc: float) → float

Returns the sep parameter needed for the target AUC for the given distribution.

DylMath.genX0X1(predicted: tuple, D1: tuple = None, D0: tuple = None) → Tuple[list, list]

Generates x0 and x1 vectors out of the given parameters.

D1 and D0 should never be smaller than the predicted array, but are often bigger.

DylMath.graphROC(predicted: tuple, D0: list = None, D1: list = None)

Generates and graphs a single ROC curve and displays the results.

DylMath.graphROCs(arrays: list, withPatches: bool = False, withLine: bool = True, D0: list = None, D1: list = None)

Graphs a collection of array predictions. Takes the arrays as they would come out of DylSort sorts.

If withPatches, puts a color coded success matrix behind the line. If withLine, graphs the line. Returns the plt handle, does not display the results.

DylMath.hanleyMcNeil(auc: float, n0: int, n1: int) → float

The very good power-law variance estimate from Hanley/McNeil.

DylMath.paramToParams(predicted: list, D0: list = None, D1: list = None) → Tuple[list, list, list]

Takes one parameter and splits it into three if predicted is a 2d list.

DylMath.runStats(groups: list, params: list, comp) → list

Runs stats on the groups provided.

Params parameter must be: ((d0d1), dist, targetAUC, n, currLayer, len(mergers))

DylMath.successMatrix(predicted: list, D0: list, D1: list)

Creates the success matrix for the predicted ordering.

Checks to make sure it got every entry filled.