utilities.py

import numpy as np
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.decomposition import PCA
from sklearn.model_selection import KFold
from sklearn.metrics import accuracy_score

from estimators import *



def get_minibatches(X,y,mb_size=20):
    n_samples = X.shape[0]
    n_batches = np.ceil(1.*n_samples/mb_size).astype(np.int32)
    all_indices = np.arange(n_samples)
    np.random.shuffle(all_indices)
    for b in range(n_batches):
        left_i  = b*mb_size
        right_i = min((b+1)*mb_size, n_samples)
        indices = sorted(all_indices[left_i:right_i])
        yield X[indices, :], y[indices][:,None]


def standard_scale_dataset(training):

    scaler = StandardScaler(copy=True, with_mean=False, with_std=False)
    trainArray = np.array(training)
    trainData, trainTarget = trainArray[:, :(len(trainArray[0]) - 1)], trainArray[:, len(trainArray[0]) - 1]
    scaler.fit(trainData)
    trainData = scaler.transform(trainData)
    trainArray = np.c_[trainData, trainTarget]
    training = trainArray.tolist()
    for item in training:
        item[-1] = int(item[-1])
    return training, scaler

# getting the scaled train and test sets and the test scaler
# training was fit and transformed the test only fit
def get_minmax_scaled_dataset_and_scaler(training):

    # instantiate the MinMax instance for training and for test
    scaler = MinMaxScaler(feature_range=(-1,1))
    trainArray = np.array(training)
    trainData, trainTarget = trainArray[:, :(len(trainArray[0]) - 1)], trainArray[:, len(trainArray[0]) - 1]
    scaler.fit(trainData)
    trainData = scaler.transform(trainData)
    trainArray = np.c_[trainData, trainTarget]
    training = trainArray.tolist()
    for item in training:
        item[-1] = int(item[-1])
    return training, scaler

def get_LDA (training):

    trainArray = np.array(training)
    xTrain = trainArray[:, :(len(trainArray[0]) - 1)]
    yTrain = trainArray[:, len(trainArray[0]) - 1]
    myPCA = PCA()
    xTrain = myPCA.fit_transform(xTrain, yTrain)
    return np.c_[xTrain, yTrain]