Note

This documentation is for a development version. Click here for the latest stable release (v1.3.0).

# nengo_spa.vector_generation¶

Generators to create vectors with specific properties.

Classes

 Generator for axis aligned vectors. Generator for equally spaced positive unitary HRR vectors. ExpectedUnitLengthVectors(d[, rng]) Generator for vectors with expected unit-length. UnitLengthVectors(d[, rng]) Generator for uniformly distributed unit-length vectors. UnitaryVectors(d, algebra[, rng]) Generator for unitary vectors (given some binding method). OrthonormalVectors(d[, rng]) Generator for random orthonormal vectors. VectorsWithProperties(d, properties, algebra, *) Generator for vectors with given properties.
nengo_spa.vector_generation.AxisAlignedVectors(d)[source]

Generator for axis aligned vectors.

Can yield at most d vectors.

Note that while axis-aligned vectors can be useful for debugging, they will not work well with most binding methods for Semantic Pointers.

Parameters

d (int) – Dimensionality of returned vectors.

Examples

>>> for p in nengo_spa.vector_generation.AxisAlignedVectors(4):
...    print(p)
[1. 0. 0. 0.]
[0. 1. 0. 0.]
[0. 0. 1. 0.]
[0. 0. 0. 1.]

class nengo_spa.vector_generation.UnitLengthVectors(d, rng=None)[source]

Bases: object

Generator for uniformly distributed unit-length vectors.

Parameters
• d (int) – Dimensionality of returned vectors.

• rng (numpy.random.RandomState, optional) – The random number generator to use to create new vectors.

class nengo_spa.vector_generation.UnitaryVectors(d, algebra, rng=None)[source]

Bases: object

Generator for unitary vectors (given some binding method).

Parameters
• d (int) – Dimensionality of returned vectors.

• algebra (AbstractAlgebra) – Algebra that defines what vectors are unitary.

• rng (numpy.random.RandomState, optional) – The random number generator to use to create new vectors.

class nengo_spa.vector_generation.EquallySpacedPositiveUnitaryHrrVectors(*, d, n, offset)[source]

Bases: object

Generator for equally spaced positive unitary HRR vectors.

The vectors produced by this generator lie all on a hyper-circle of positive, unitary vectors under the HrrAlgebra. The distance from one vector to the next is constant.

Note that the identity vector is included in the set of returned vectors if any of the vectors hits an offset of 0. This might not be desired as it will return any vector it is bound to unchanged. Use a non-integer offset to ensure that the identity vector is not included.

Parameters
• d (int) – Dimensionality of returned vectors.

• n (int) – Number of vectors to fit onto the hyper-circle. At most n vectors can be returned from the generator.

• offset (float) – Offset of the first returned vector along the hyper-circle. An offset of 0 will return the identity vector first. An offset of 1 corresponds to the vector when moving a 1/n-th part along the hyper-circle.

vectors

All vectors that would be returned by iterating over the generator.

Type

(n, d) ndarray

class nengo_spa.vector_generation.OrthonormalVectors(d, rng=None)[source]

Bases: object

Generator for random orthonormal vectors.

Parameters
• d (int) – Dimensionality of returned vectors.

• rng (numpy.random.RandomState, optional) – The random number generator to use to create new vectors.

class nengo_spa.vector_generation.ExpectedUnitLengthVectors(d, rng=None)[source]

Bases: object

Generator for vectors with expected unit-length.

The vectors will be uniformly distributed with an expected norm of 1, but each specific pointer may have a length different than 1. Specifically each vector component will be normal distributed with mean 0 and standard deviation $$1/\sqrt{d}$$.

Parameters
• d (int) – Dimensionality of returned vectors.

• rng (numpy.random.RandomState, optional) – The random number generator to use to create new vectors.

class nengo_spa.vector_generation.VectorsWithProperties(d, properties, algebra, *, rng=None)[source]

Bases: object

Generator for vectors with given properties.

Supported properties depend on the algebra. See the respective algebra’s AbstractAlgebra.create_vector() method.

Parameters
• d (int) – Dimensionality of returned vectors.

• properties – Properties that the generated vectors have to fulfill. Details depend on the exact algebra.

• algebra (AbstractAlgebra) – Algebra that determines the interpretation of the properties.

• rng (numpy.random.RandomState, optional) – The random number generator to use to create new vectors.