helperfns 1.0.2

This package provide some python helper functions that are useful in machine learning.

helperfns

🎀 This is a python package that contains some helper functions for machine leaning.

---

Table of Contents

• helperfns
• Table of Contents
• Getting started
• Usage
• tables
• text
• utils
• visualization
• Contributing to helperfns.
• Documentation
• License

Getting started

To start using helperfns in your project you run the following command:

pip install helperfns

Or if you wan to install it in notebooks such as jupyter notebooks you can run the code cell with the following code:

!pip install helperfns

Usage

The helperfns package is made up of different sub packages such as:

1. tables
2. text
3. utils
4. visualization

tables

In the tables sub package you can print your data in tabular form for example:

from helperfns.tables import tabulate_data

column_names = ["SUBSET", "EXAMPLE(s)", "Hello"]
row_data = [["training", 5, 4],['validation', 4, 4],['test', 3, '']]
tabulate_data(column_names, row_data)

Output:

Table
+------------+------------+-------+
| SUBSET     | EXAMPLE(s) | Hello |
+------------+------------+-------+
| training   |          5 |     4 |
| validation |          4 |     4 |
| test       |          3 |       |
+------------+------------+-------+

The following is the table of arguments for the tabulate_data helper function

Argument	Description	Type
column_names	List of column names	list
data	Data to be tabulated	list
title	Title of the table	str

text

The text package offers two main function which are clean_sentence, de_contract, generate_ngrams and generate_bigrams

from helperfns.text import *

# cleans the sentence
print(clean_sentence("text 1 # https://url.com/bla1/blah1/"))

Here is the table of arguments for the clean_sentence helper function.

Argument	Description	Type
sent	Input sentence	str
lower	Flag to convert to lower case (default: True)	bool

You can get the list of english words as follows:

# list of all english words
print(english_words)

You can use the de_contract to de-contact strings as follows

# converts strings like `I'm` to 'I am'
print(de_contract("I'm"))

Here is the table of arguments for the de_contract function.

Argument	Description	Type
word	Word to de-contract	str

The generate_bigrams is responsible for generating bi grams from list of words. Here is how you can use the function

# generate bigrams from a list of word
print(text.generate_bigrams(['This', 'film', 'is', 'terrible']))

Here is the table of arguments for the generate_bigrams function:

Argument	Description	Type
x	List of input elements	list

The generate_ngrams generate the n-grams from a list of words, here is an example on how you can use this function

# generates n-grams from a list of words
print(text.generate_ngrams(['This', 'film', 'is', 'terrible']))

Here is the table of arguments for the generate_ngrams function:

Argument	Description	Type
x	List of input elements	list
grams	Number of grams for generating n-grams (default: 3)	int

utils

utils package comes with a simple helper function for converting seconds to hours, minutes and seconds.

Example:

from helperfns.utils import hms_string

start = time.time()
for i in range(100000):
   pass
end = time.time()

print(hms_string(end - start))

Output:

'0:00:00.01'

The hms_string takes in the following as arguments.

Argument	Description	Type
sec_elapsed	Time in seconds to be converted	Any

visualization

This sub package provides different helper functions for visualizing data using plots.

Examples:

The following code cell will plot a classification report of true labels versus predicted labels.

from helperfns.visualization import plot_complicated_confusion_matrix, plot_images, plot_images_predictions, plot_simple_confusion_matrix,
plot_classification_report

# plotting classification report
fig, ax = plot_classification_report(labels, preds,
                    title='Classification Report',
                    figsize=(10, 5), dpi=70,
                    target_names = classes)

The plot_classification_report takes the following arguments:

Argument	Description	Type
y_true	True labels	list
y_pred	Predicted labels	list
title	Title of the plot (default: "Classification Report")	str
figsize	Size of the figure (default: (10, 5))	tuple
dpi	Resolution of the figure (default: 70)	int
save_fig_path	Path to save the figure (default: None)	Any or None
**kwargs	Additional keyword arguments	Any

The plot_images_predictions plots the image predictions. This functions is very useful when you are doing image classification.

# plot predicted image labels with the images
plot_images_predictions(images, true_labels, preds, classes=["dog", "cat"] ,cols=8)

Here is the table of arguments for the plot_images_predictions.

Argument	Description	Type
images	List of images to plot	list
labels_true	True labels	list
labels_pred	Predicted labels	list
classes	List of class labels (default: [])	list
cols	Number of columns in the plot (default: 5)	int
rows	Number of rows in the plot (default: 3)	int
fontsize	Font size for labels (default: 16)	int

The plot_images functions is used to visualize images.

# plot the images with their labels
plot_images(images[:24], true_labels[:24], cols=8)

The plot_images takes the following as arguments:

Argument	Description	Type
images	List of images to plot	list
labels	List of labels corresponding to images	list
cols	Number of columns in the plot (default: 5)	int
rows	Number of rows in the plot (default: 3)	int
fontsize	Font size for labels (default: 16)	int

The plot_simple_confusion_matrix is used to plot a less more verbose confusion matrix of real labels against predicted labels.

# plot a simple confusion matrix
y_true = [random.randint(0, 1) for _ in range (100)]
y_pred = [random.randint(0, 1) for _ in range (100)]
classes =["dog", "cat"]
plot_simple_confusion_matrix(y_true, y_pred, classes)

This function takes in the following in the following as arguments.

Argument	Description	Type
y_true	True labels	list
y_pred	Predicted labels	list
classes	List of class labels (default: [])	list
figsize	Size of the figure (default: (10, 10))	tuple
fontsize	Font size for labels (default: 15)	int

The plot_complicated_confusion_matrix is used to plot a more verbose confusion matrix of real labels against predicted labels.

# plot a confusion matrix with percentage value of confusion
y_true = [random.randint(0, 1) for _ in range (100)]
y_pred = [random.randint(0, 1) for _ in range (100)]
classes =["dog", "cat"]
plot_complicated_confusion_matrix(y_true, y_pred, classes)

This function takes in the following as arguments.

Argument	Description	Type
y_true	True labels	list
y_pred	Predicted labels	list
classes	List of class labels (default: [])	list
figsize	Size of the figure (default: (5, 5))	tuple
fontsize	Font size for labels (default: 20)	int
title	Title of the plot (default: "Confusion Matrix")	str
xlabel	Label for x-axis (default: "Predicted label")	str
ylabel	Label for y-axis (default: "True label")	str

Contributing to helperfns.

To contribute to helperfns read the CONTRIBUTION.md file.

Documentation

You can read the full documentation here.

License

This project is licensed under the MIT License - see the LICENSE file for details.