twolodzko

# Code for:
# https://stats.stackexchange.com/questions/360863/probability-based-on-observed-data/360892#360892

set.seed(123)

A <- c(34.5, 34.2, 35, 35.1, 36, 35.2, 35.7, 34.8, 34.9, 34.4)
B <- c(35, 34.2, 36, 35, 34, 34.2, 34.2, 34.5, 34.4)

twolodzko

twolodzko

from tqdm import tqdm

def tqdm_function_decorator(*args, **kwargs):
    """
    Decorate a function by adding a progress bar

    Parameters
    ----------
    *args, **kwargs

twolodzko

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from six import iteritems

from tqdm import tqdm
from collections import Counter
import re

class TopKTokenizer(object):

twolodzko

# Many of the functions are adapted from:
# https://www.polydesmida.info/cookbook/functions.html


nohead ()
{
if [[ $# -eq 0 || "$1" == "-h" ]] ; then
  echo "Useage: nohead file [n=1]"
  echo "Omit header (n rows) from a file."

twolodzko

import xxhash
from keras.preprocessing.text import hashing_trick

# one_hot and hashing_trick in Keras both use by default python's hash function
# it is unstable: https://stackoverflow.com/q/27522626/3986320
# alternatively, you could use md5, but it's not the fastest hashing function
# xxHash package offers a faster alternative

xxh = lambda w: int(xxhash.xxh32(w.encode()).hexdigest(), 16)
one_hot = lambda x, n, **kwargs: hashing_trick(x, n, hash_function=xxh, **kwargs)

twolodzko

from keras.callbacks import Callback
import matplotlib.pyplot as plt

class LRFinder(Callback):
    
    '''
    A simple callback for finding the optimal learning rate range for your model + dataset. 
    
    # Usage
        ```python

twolodzko

from keras import backend as K
from keras.layers import Average
from keras.activations import softmax

class WeightedAverage(Average):
  
  def build(self, input_shape):
    self.kernel = self.add_weight(name='kernel', 
                                  shape=(1, len(input_shape)),
                                  initializer='ones',

twolodzko

import numpy as np

def train_test_split(*arrays, test_size, random_state, clusters):
  '''Split to train and test samples by clusters
  
  Parameters
  ----------
  
  test_size : float, 0 < test_size < 1
    fraction of clusters to include in test set

twolodzko

import numpy as np
from sklearn.metrics import roc_curve

def sensivity_specifity_cutoff(y_true, y_score):
    '''Find data-driven cut-off for classification
    
    Cut-off is determied using Youden's index defined as sensitivity + specificity - 1.
    
    Parameters
    ----------