block_vector

Implementation of a general “block vector”

The pyomo.contrib.pynumero.sparse.block_vector module includes methods that extend linear algebra operations in numpy for case of structured problems where linear algebra operations present an inherent block structure. This interface consider vectors of the form:

\[v = [v_1, v_2, v_3, ... , v_n]\]

where v_i are numpy arrays of dimension 1

Contents

Methods specific to BlockVector:

• set_block()

• get_block()

• block_sizes()

• get_block_size()

• is_block_defined()

• copyfrom()

• copyto()

• copy_structure()

• set_blocks()

• pprint()

Attributes specific to BlockVector:

• nblocks

• bshape

• has_none

NumPy compatible methods:

• dot()

• sum()

• all()

• any()

• max()

• astype()

• clip()

• compress()

• conj()

• conjugate()

• nonzero()

• ptp() (NumPy 1.x only)

• round()

• std()

• var()

• tofile()

• min()

• mean()

• prod()

• fill()

• tolist()

• flatten()

• ravel()

• argmax()

• argmin()

• cumprod()

• cumsum()

• copy()

For example,

>>> import numpy as np
>>> from pyomo.contrib.pynumero.sparse import BlockVector
>>> v = BlockVector(2)
>>> v.set_block(0, np.random.normal(size=100))
>>> v.set_block(1, np.random.normal(size=30))
>>> avg = v.mean()

NumPy compatible functions:

• log10()

• sin()

• cos()

• exp()

• ceil()

• floor()

• tan()

• arctan()

• arcsin()

• arccos()

• sinh()

• cosh()

• abs()

• tanh()

• arccosh()

• arcsinh()

• arctanh()

• fabs()

• sqrt()

• log()

• log2()

• absolute()

• isfinite()

• isinf()

• isnan()

• log1p()

• logical_not()

• expm1()

• exp2()

• sign()

• rint()

• square()

• positive()

• negative()

• rad2deg()

• deg2rad()

• conjugate()

• reciprocal()

• signbit()

• add()

• multiply()

• divide()

• subtract()

• greater()

• greater_equal()

• less()

• less_equal()

• not_equal()

• maximum()

• minimum()

• fmax()

• fmin()

• equal()

• logical_and()

• logical_or()

• logical_xor()

• logaddexp()

• logaddexp2()

• remainder()

• heaviside()

• hypot()

For example,

>>> import numpy as np
>>> from pyomo.contrib.pynumero.sparse import BlockVector
>>> v = BlockVector(2)
>>> v.set_block(0, np.random.normal(size=100))
>>> v.set_block(1, np.random.normal(size=30))
>>> inf_norm = np.max(np.abs(v))

BlockVector(nblocks)	Structured vector interface.
BlockVector.set_block(key, value)	Set a block.
BlockVector.get_block(key)	Access a block.
BlockVector.block_sizes([copy])	Returns 1D-Array with sizes of individual blocks in this BlockVector
BlockVector.get_block_size(ndx)
BlockVector.is_block_defined(ndx)
BlockVector.copyfrom(other)	Copy entries of other vector into this vector
BlockVector.copyto(other)	Copy entries of this BlockVector into other
BlockVector.copy_structure()	Returns a copy of the BlockVector structure filled with zeros
BlockVector.set_blocks(blocks)	Assigns vectors in blocks
BlockVector.pprint()	Prints BlockVector in pretty format
BlockVector.nblocks	Returns the number of blocks.
BlockVector.bshape	Returns the number of blocks in this BlockVector in a tuple.
BlockVector.has_none	Indicate if this BlockVector has any none entries.

Classes

BlockVector(nblocks)

Structured vector interface.

Exceptions

NotFullyDefinedBlockVectorError

Functions

assert_block_structure(vec)