decogo.util.block_vector¶

Implements sparse and dense block vector data structures

Functions

infinity_norm(x)

Computes infinity norm, i.e. \(||x||_\infty = \max\limits_{k \in K, i \in n_k}|x_{ki}|\).

Classes

`BlockVector`([blocks_size])	This class stores dense vector with the respect to blocks
`SparseBlockVector`(values, blocks_size)	Class for storing the values in sparse format blockwise.

class decogo.util.block_vector.SparseBlockVector(values, blocks_size)[source]¶

Class for storing the values in sparse format blockwise. It consists of k blocks. Each block uses the standard Python sparse matrix with dimension \(1 \times n_k\), where \(n_k\) is the block size

Parameters:

vectors (dict) – Sparse matrices stored blockwise
blocks_size (list) – Size of the blocks
blocks (list) – Block ids which contain nonzero values

__init__(values, blocks_size)[source]¶

Constructor method

Parameters:

values (list) – List of tuples (block_id, index, value)
blocks_size (list) – Number of variables per block

__getitem__(item)[source]¶

Index operator

Parameters:: item (tuple) – Input index as tuple: (block_id, index)
Returns:: Value associated with the given index
Return type:: float

scalar_prod(x)[source]¶

Scalar product computation (or dot product) computed blockwise, \(v=\sum_{k \in K}a_k^Tx_k\)

Parameters:: x (SparseBlockVector or BlockVector) – Input vector
Returns:: Value for scalar product \(v\)
Return type:: float

maxabs(block_id=None)[source]¶

Finds maximum element by absolute value

Parameters:: block_id (int, optional) – Block id where to look for the maximum value
Returns:: Max element by absolute value
Return type:: float or None

property num_blocks¶

Return number of blocks

Returns:: Return number of blocks
Return type:: int

get_block(block_id)[source]¶

Gets vector associated with the given block id

Parameters:: block_id (int) – Given block identifier
Returns:: Sparse vector (matrix)
Return type:: coo_matrix or None

get_block_size(k)[source]¶

Get block size associated with the given block id

Parameters:: k (int) – Given block identifier
Returns:: Block size
Return type:: int

get_norm()[source]¶

class decogo.util.block_vector.BlockVector(blocks_size=None)[source]¶

This class stores dense vector with the respect to blocks

Parameters:

vectors (dict) – Dense vectors stored blockwise
blocks_size (list) – Size of the blocks
blocks (list) – Block identifiers

__init__(blocks_size=None)[source]¶: Constructor method

__getitem__(item)[source]¶

Index operator

Parameters:: item (tuple) – Input index in tuple (block_id, index)
Returns:: Value associated to index
Return type:: float

get_block(block_id)[source]¶

Get vector associated to the blocks

Parameters:: block_id – Block identifier
Returns:: Vector
Return type:: ndarray

set_block(block_id, array)[source]¶

Sets vector associated to block

Parameters:

block_id (int) – Block identifier
array (ndarray) – Array to set

__setitem__(key, value)[source]¶

Index operator

Parameters:

key (tuple) – Index as tuple: (block_id, index)
value (float) – Value to set

__str__()[source]¶

String operator

Returns:: Nicer way of string representation
Return type:: str

property num_blocks¶

Return number of blocks

Returns:: Number of blocks
Return type:: int

__add__(other)[source]¶

Operator +

Parameters:: other (BlockVector) – Other vector
Returns:: Sum of two vectors
Return type:: BlockVector

__sub__(other)[source]¶

Operator -

Parameters:: other (BlockVector) – Other vector
Returns:: Difference of two vectors
Return type:: BlockVector

__eq__(other)[source]¶

Operator ==, checks if two vectors are equal

Parameters:: other (BlockVector) – Other vector
Returns:: True or False
Return type:: bool

__mul__(other)[source]¶

Multiplication operator. Currently supports only scalar types for multiplication

Parameters:: other (float) – Scalar value
Returns:: Vector multiplied by scalar value
Return type:: BlockVector
Raises:: ValueError – If other is not a sequence

__truediv__(other)[source]¶

True division operator. Currently supports only scalar types for division

Parameters:: other (float) – Scalar value
Returns:: Vector divided by scalar value
Return type:: BlockVector
Raises:: ValueError – If other is not a sequence

__abs__()[source]¶

Absolute value operator

Returns:: New vector with absolute values
Return type:: BlockVector

decogo.util.block_vector.infinity_norm(x)[source]¶

Computes infinity norm, i.e. \(||x||_\infty = \max\limits_{k \in K, i \in n_k}|x_{ki}|\)

Parameters:: x (BlockVector or SparseBlockVector) – Input vector
Returns:: Value of infinity norm
Return type:: float