SyntheticGen 0.1.0

Synthetic dataset creator/augmentor for machine learning applications.

SyntheticGen v0.1.1

This is a Python package available on PyPi that creates synthetic datasets for various dataset types.

SyntheticGen currently supports:

• Linear Synthetic Datasets

• Linear Augmentations

Installation

pip install syntheticgen

Usage

In v0.1.1 there is only a linear augmentor available however, image and experimental augmentation techniques will be available soon.

#Import a dataset as a numpy array (in this example it is named dataset)

#Create an augmentor object and initialize it with the dataset

aug = linearAugmentor(dataset)





#addMatrix() and setPowerRange() will configure the augmentor with custom specifications.

aug.addMatrix(.25)

aug.setPowerRange(2,2)



#performOperations() will run the augmentation operations with the set configuration.

aug.performOperation()



#Returns the augmented dataset

dataset = aug.getCombinedSet()

Object and Method Details

linearAugmentor(dataset):\

This is the augmentor object for psuedo-linear datasets.

Dataset Specifications For Proper Function:

• Must be a numpy array with sub-lists holding inputs and outputs in the same order every line.

• Each sub-list in the dataset must be of equal length to one another

linearAugmentor.addMatrix(percentAdded):\

This method will randomly add a custom percentage of the inputed dataset into a matrix to then be augmented.

Parameters:

• percentAdded

  • This is a float value that determines the percentage of the inputted dataset that will be augmented.

  • Minimum to function: whatever float equates to 2 lines of the dataset.

  • Maximum is 1.0, the entire dataset.

linearAugmentor.setIntRange(lowerBound, upperBound):\

This method will set an integer range for the number of operations to be performed on the matrix.

Parameters:

• lowerBound

  • This is the fewest possible operations to be performed on the inputted dataset's matrix.

  • Minimum: 0

  • No Maximum

• upperBound

  • This is the maximum number of possible operations to be performed on the inputted dataset's matrix.

  • Minimum: 0

  • No Maximum

linearAugmentor.setPowerRange(lowerBound, upperBound):\

This method will set a range for the number of operations to be performed on the matrix based on the length of the matrix raised to a power (the bounds).

Parameters:

• lowerBound

  • This is the value the length of the matrix will be raised by creating the fewest possible operations to be performed.

  • Minimum: 0

  • No Maximum

• upperBound

  • This is the value the length of the matrix will be raised by creating the maximum number of possible operations to be performed.

  • Minimum: 0

  • No Maximum

linearAugmentor.performOperations():\

This is the method that will perform the operations on the matrix given the inputted specifications.

linearAugmentor.getCombinedSet():\

This method returns the original dataset randomly combined with the newly created augmented data.

linearAugmentor.getSyntheticData()\

This method returns only the augmented data creating a fully synthetic dataset.

linearAugmentor.getInitialDataset()\

This method returns the originally inputted, unchanged dataset.

Roadmap

Linear Augmentor

• Add additional customization options pertaining to the type of operations performed

• Add more advanced augmentation techniques utilizing eingenvectors.

Image Augmentor

• Add an image augmentor that performs multiple augmentation techniques.

Differential Augmentor

• Experimental augmentor that's currently in production utilizing differential calculus.

Recent Changes

v0.1.1

• README changes.

v0.1.0

• Created package and added the linear augmentor.