featurevectormatrix package

Module contents

class featurevectormatrix.FeatureVectorMatrix(default_value=0, default_to_hashed_rows=False, rows=None)

Bases: object

A class to abstract away the differences in internal representation between dictionaries and lists that can matter for very large datasets of vectors and allow them to work seamlessly with each other

Supports indexing and iteration (fvm[1] and for i in fvm:...) but you should set default_to_hash_rows to get the expected behavior. Also supports len

add_row(list_or_dict, key=None)

Adds a list or dict as a row in the FVM data structure

Parameters:

• key (str) – key used when rows is a dict rather than an array
• list_or_dict – a feature list or dict

column_count()

Get the current number of columns

Returns:the count

column_names()

get the column names

Returns:The ordered list of column names

default_to_hashed_rows(default=None)

Gets the current setting with no parameters, sets it if a boolean is passed in

Parameters:default – the value to set Returns:the current value, or new value if default is set to True or False

extend_rows(list_or_dict)

Add multiple rows at once

Parameters:list_or_dict – a 2 dimensional structure for adding multiple rows at once Returns:

get_matrix()

Use numpy to create a real matrix object from the data

Returns:the matrix representation of the fvm

get_row_dict(row_idx)

Return a dictionary representation for a matrix row

Parameters:row_idx – which row Returns:a dict of feature keys/values, not including ones which are the default value

get_row_list(row_idx)

get a feature vector for the nth row

Parameters:row_idx – which row Returns:a list of feature values, ordered by column_names

keys()

Returns all row keys

Raises NotImplementedError:  if all rows aren’t keyed Returns:all row keys

row_count()

The current number of rows

Returns:the count

row_names()

get the column names

Returns:The ordered list of column names

set_column_names(column_names)

Setup the feature vector with some column names :param column_names: the column names we want :return:

set_row_names(row_names)

Setup the feature vector with some column names :param row_names: the column names we want :return:

transpose()

Create a matrix, transpose it, and then create a new FVM

Raises NotImplementedError:  if all existing rows aren’t keyed Returns:a new FVM rotated from self