Ángel Igareta angeligareta

## stratified_sampling.ipynb

      
              1 file
            
          
              0 forks
            
          
              0 comments
            
          
              6 stars
            
          
                angeligareta
                / stratified_sampling.ipynb
            
            
              Created
              July 21, 2021 06:46
            
          
      Sorry, something went wrong. Reload?
      Sorry, we cannot display this file.
      Sorry, this file is invalid so it cannot be displayed.
      
          Viewer requires iframe.
      
    
## dhash_z_transformed.py
# Inspired from https://github.com/JohannesBuchner/imagehash repository
import imagehash
from PIL import Image

def alpharemover(image):
    if image.mode != 'RGBA':
        return image
    canvas = Image.new('RGBA', image.size, (255,255,255,255))
    canvas.paste(image, mask=image)
    return canvas.convert('RGB')

## custom_early_stopping.py
import tensorflow as tf
import numpy as np


class CustomEarlyStopping(tf.keras.callbacks.Callback):
    """
    Custom Early Stopping callback to monitor multiple metrics by combining them using a harmonic mean calculation.

    Adapted from (TensorFlow EarlyStopping source)[https://github.com/tensorflow/tensorflow/blob/v2.5.0/tensorflow/python/keras/callbacks.py#L1683-L1823].

## generate_class_weights.py
import numpy as np
from sklearn.utils.class_weight import compute_class_weight
from sklearn.preprocessing import MultiLabelBinarizer


def generate_class_weights(class_series, multi_class=True, one_hot_encoded=False):
  """
  Method to generate class weights given a set of multi-class or multi-label labels, both one-hot-encoded or not.

  Some examples of different formats of class_series and their outputs are:

## get_dataset_partitions_pd.py
def get_dataset_partitions_pd(df, train_split=0.8, val_split=0.1, test_split=0.1):
    assert (train_split + test_split + val_split) == 1

    # Only allows for equal validation and test splits
    assert val_split == test_split

    # Specify seed to always have the same split distribution between runs
    df_sample = df.sample(frac=1, random_state=12)
    indices_or_sections = [int(train_split * len(df)), int((1 - val_split - test_split) * len(df))]


## get_dataset_partitions_tf.py
def get_dataset_partitions_tf(ds, ds_size, train_split=0.8, val_split=0.1, test_split=0.1, shuffle=True, shuffle_size=10000):
    assert (train_split + test_split + val_split) == 1

    if shuffle:
        # Specify seed to always have the same split distribution between runs
        ds = ds.shuffle(shuffle_size, seed=12)

    train_size = int(train_split * ds_size)
    val_size = int(val_split * ds_size)


## get_best_epoch_per_model.py
import pandas as pd
import tensorboard as tb
import statistics

def get_best_epoch_per_model(
    experiment_id,
    metrics,
    format_as_percentage = False,
    sort_values = False,
    group_by_model_names = False,

## language_selector_buddypress.php
<?php
// Add custom language selector field for XProfileFields Buddypress
// Author: Angel Igareta, adapted from https://tulipemedia.com/en/buddypress-auto-bulk-import-countries-profile-field/
function bp_add_custom_language_list() {
    if ( !xprofile_get_field_id_from_name('Languages Spoken') && 'bp-profile-setup' == $_GET['page'] ) {
        $language_list_args = array(
            'field_group_id' => 1,
            'name' => 'Languages Spoken',
            'description' => 'Please select the languages you speak',
            'can_delete' => true,

## No toques aquí - Broma Whatsapp Feeling The Net
👉🏼<⚫>👈🏼 ‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎ ‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎ ‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎ ‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎ ‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎ ‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏‎‏
	# Inspired from https://github.com/JohannesBuchner/imagehash repository
	import imagehash
	from PIL import Image

	def alpharemover(image):
	if image.mode != 'RGBA':
	return image
	canvas = Image.new('RGBA', image.size, (255,255,255,255))
	canvas.paste(image, mask=image)
	return canvas.convert('RGB')
	import tensorflow as tf
	import numpy as np


	class CustomEarlyStopping(tf.keras.callbacks.Callback):
	"""
	Custom Early Stopping callback to monitor multiple metrics by combining them using a harmonic mean calculation.

	Adapted from (TensorFlow EarlyStopping source)[https://github.com/tensorflow/tensorflow/blob/v2.5.0/tensorflow/python/keras/callbacks.py#L1683-L1823].
	import numpy as np
	from sklearn.utils.class_weight import compute_class_weight
	from sklearn.preprocessing import MultiLabelBinarizer


	def generate_class_weights(class_series, multi_class=True, one_hot_encoded=False):
	"""
	Method to generate class weights given a set of multi-class or multi-label labels, both one-hot-encoded or not.

	Some examples of different formats of class_series and their outputs are:
	def get_dataset_partitions_pd(df, train_split=0.8, val_split=0.1, test_split=0.1):
	assert (train_split + test_split + val_split) == 1

	# Only allows for equal validation and test splits
	assert val_split == test_split

	# Specify seed to always have the same split distribution between runs
	df_sample = df.sample(frac=1, random_state=12)
	indices_or_sections = [int(train_split * len(df)), int((1 - val_split - test_split) * len(df))]
	def get_dataset_partitions_tf(ds, ds_size, train_split=0.8, val_split=0.1, test_split=0.1, shuffle=True, shuffle_size=10000):
	assert (train_split + test_split + val_split) == 1

	if shuffle:
	# Specify seed to always have the same split distribution between runs
	ds = ds.shuffle(shuffle_size, seed=12)

	train_size = int(train_split * ds_size)
	val_size = int(val_split * ds_size)
	import pandas as pd
	import tensorboard as tb
	import statistics

	def get_best_epoch_per_model(
	experiment_id,
	metrics,
	format_as_percentage = False,
	sort_values = False,
	group_by_model_names = False,
	<?php
	// Add custom language selector field for XProfileFields Buddypress
	// Author: Angel Igareta, adapted from https://tulipemedia.com/en/buddypress-auto-bulk-import-countries-profile-field/
	function bp_add_custom_language_list() {
	if ( !xprofile_get_field_id_from_name('Languages Spoken') && 'bp-profile-setup' == $_GET['page'] ) {
	$language_list_args = array(
	'field_group_id' => 1,
	'name' => 'Languages Spoken',
	'description' => 'Please select the languages you speak',
	'can_delete' => true,