Data

class metamorphic_relations.Data.Data(train_x, train_y, test_x, test_y, max_y)

Stores the data

Parameters:

• train_x (array) – a numpy array, the first dimension is the index of elements for training

• train_y (array) – a 1D numpy array of label indices for training

• test_x (array) – a numpy array, the first dimension is the index of elements for testing

• test_y (array) – a 1D numpy array of label indices for testing

• max_y (int) – the largest y value possible

static concat_lists(lists)

Takes a list of pairs of numpy arrays of xs and ys and makes them a single xs and ys list

Parameters:

lists (list[tuple[numpy.array, numpy.array]]) – a numpy array of pairs of numpy arrays of xs and ys e.g. [[xs1, ys1], [xs2, ys2]]

Returns:

a tuple of xs and ys e.g. (xs1 + xs2, ys1 + ys2)

Return type:

tuple[numpy.array, numpy.array]

get_train_subset(i_min=0, i_max=9999999)

Gets a subset of the training data

Parameters:

• i_min (int) – the lower bound index (inclusive)

• i_max (int) – the upper bound index (not inclusive)

Returns:

train_x_subset, train_y_subset

Return type:

tuple[numpy.array, numpy.array]

static group_by_label(y, max_y)

Groups an array of ints by their values. E.g. ([3, 3, 2, 3, 1, 0], 5) -> [[5], [4], [2], [0, 1, 3], []]

Parameters:

• y (array) – a numpy array of ints

• max_y (int) – the maximum possible number of values

Returns:

a list of y indices for each possible y value

Return type:

list[list[int]]

shuffle_train()

Rearranges the order of the training data

update_test(test_x, test_y)

Updates the testing data

Parameters:

• test_x (array) – new testing x data

• test_y (array) – new testing y data

update_train(train_x, train_y)

Updates the training data

Parameters:

• train_x (array) – new training x data

• train_y (array) – new training y data

class metamorphic_relations.Results.Results(original_results=None, GMR_results=None, DSMR_results=None, all_MR_results=None, individual_results=None)

Create an object to store and manipulate the results given multiple sets of metamorphic relations (MRs)

Parameters:

• original_results (Info) – the results using the original data

• GMR_results (Info) – the results when augmenting the data with generic MRs (GMRs)

• DSMR_results (Info) – the results when augmenting the data with domain specific MRs (DSMRs)

• all_MR_results (Info) – the results when augmenting the data with GMRs and DSMRs

• individual_results (list[Info]) – a list of results from individual MRs

static get_JSON(info=None)

Parameters:

info (Info | list[Info] | None) – the Info or list of Info objects to convert

Returns:

None or a string in JSON form for the object

Return type:

str

get_forall_sets(is_set=(True, True, True, True), get_set_function=None)

For all sets of results which are not None call a function

Parameters:

• is_set (tuple[bool, bool, bool, bool]) – the sets of results to use of [“original_results”, “GMR_results”, “DSMR_results”, “all_MR_results”]. E.g. [True, False, True, False] uses [“original_results”, “DSMR_results”]

• get_set_function (function) – the function to be used with the result sets

Returns:

the results of the function for each non None set

Return type:

list

graph(set_name='', train_f1s=False, test_f1s=True, original_counts=True, show_sets=(True, True, True, True))

Graphs the results of the deep learning with MRs

Parameters:

• set_name (str) – the name of the set trained, to be used in the title

• train_f1s (bool) – choose whether to show the train F1 scores

• test_f1s (bool) – choose whether to show the test F1 scores

• original_counts (bool) – choose whether to show the F1 scores against the number of original training elements or the actual counts (number of training elements after the MRs)

• show_sets (tuple[bool, bool, bool, bool]) – the sets of results to show of [“original_results”, “GMR_results”, “DSMR_results”, “all_MR_results”]. E.g. [True, False, True, False] shows [“original_results”, “DSMR_results”]

graph_all(set_name='')

Graphs the train and test results using original and actual counts

Parameters:

set_name (str) – the name of the set trained, to be used in all the graph titles

print_individual()

Prints the individual results as a table

static read_from_file(filename)

Reads results from a file in the Results class json format

Parameters:

filename (str) – the file (or path) to read the data from

Returns:

a Results object

Return type:

Results

write_to_file(filename)

Writes the results to a file

Parameters:

filename (str) – the file (or path) to write the data to

Returns:

a string representation of the Results object

Return type:

str

class metamorphic_relations.Info.Info(original_count, actual_count, train_f1, test_f1, name=None)

Create an object to store and manipulate the information for a given set of metamorphic relations (MRs)

Parameters:

• original_count (list[int]) – the number of training elements before the MRs were used

• actual_count (list[int]) – the number of training elements used in training (i.e. after the MRs were used)

• train_f1 (list[float]) – the average F1 score over the training folds

• test_f1 (list[float]) – the F1 score calculated on the test set

• name (str) – the name of the info stored

static from_JSON(string)

Turns a JSON to an Info object

Parameters:

string (str) – a string in JSON form with keys: “actual_count”, “original_count”, “train_f1”, “test_f1”, “name”

Returns:

an Info object

Return type:

Info

static list_to_info(results)

Turns a list of results to an Info object

Parameters:

results (list[tuple[int, int, float, float]]) – a list of tuples containing (original_count, actual_count, train_f1, test_f1)

Returns:

an Info object

Return type:

Info

set_name(name)

Names this Info object

Parameters:

name (str) – the name to set

to_JSON()

Turns this object to a JSON

Returns:

string in JSON format

Return type:

str