ConvGeNCode/library/dataset.py

import math
import numpy as np

class DataSet:
    """
    Stores data and Labels.
    """
    def __init__(self, data0=None, data1=None):
        self.data0 = data0
        self.data1 = data1
        self.size0 = len(data0) if data0 is not None else 0
        self.size1 = len(data1) if data1 is not None else 0

        if data0 is not None and data1 is not None:
            self.data = np.concatenate( [data1, data0] )
            self.labels = np.concatenate( [self.labels1(), self.labels0()] )
        elif data0 is None:
            self.data = data1
            self.labels = self.labels1()
        elif data1 is None:
            self.data = data0
            self.labels = self.labels0()
        else:
            raise AttributeError("Expected data, data0 or data1 to be a numpy.array")

    def shuffleWith(self, shuffleFn):
        if self.data0 is not None:
            self.data0 = shuffleFn(self.data0)

        if self.data1 is not None:
            self.data1 = shuffleFn(self.data1)

        if self.data0 is None:
            self.data = self.data1
        elif self.data1 is None:
            self.data = self.data0
        else:
            self.data = np.concatenate((self.data1, self.data0))

    def labels0(self):
        return np.zeros(self.size0)

    def labels1(self):
        return np.zeros(self.size0) + 1


class TrainTestData:
    """
    Stores features, data and labels for class 0 and class 1.
    """

    def __init__(self, train, test):
        self.train = train
        self.test = test

    @staticmethod
    def splitUpData(data, trainFactor=0.9):
        size = len(data)
        trainSize = math.ceil(size * trainFactor)
        trn = data[list(range(0, trainSize))]
        tst = data[list(range(trainSize, size))]
        return trn, tst

    @classmethod
    def splitDataByFactor(cls, features0, features1, trainFactor=0.9):
        features_0_trn, features_0_tst = cls.splitUpData(features0, trainFactor)
        features_1_trn, features_1_tst = cls.splitUpData(features1, trainFactor)

        return cls(
            test=DataSet(data1=features_1_tst, data0=features_0_tst),
            train=DataSet(data1=features_1_trn, data0=features_0_trn)
            )

    @classmethod
    def splitDataToSlices(cls, bigData, numOfSlices=5):
        data0slices = cls._arrayToSlices(bigData.data0, numOfSlices)
        data1slices = cls._arrayToSlices(bigData.data1, numOfSlices)

        for n in range(numOfSlices):
            data0 = np.concatenate([data0slices[k] for k in range(numOfSlices) if n != k])
            data1 = np.concatenate([data1slices[k] for k in range(numOfSlices) if n != k])
            train = DataSet(data0=data0, data1=data1)
            test = DataSet(data0=data0slices[n], data1=data1slices[n])
            yield cls(train=train, test=test)

    @staticmethod
    def _arrayToSlices(data, numOfSlices):
        size = len(data)
        if size < numOfSlices:
            raise AttributeError(
                f"Expected data set to contain at least {numOfSlices} points"
                + f" but got {size} points."
                )

        sliceSize = (size // numOfSlices) + (0 if size % numOfSlices == 0 else 1)

        return [
            data[n * sliceSize : min(size, (n+1) * sliceSize)]
            for n in range(numOfSlices)
            ]