kaira.models.components.ProjectionType

Inheritance diagram for ProjectionType

class kaira.models.components.ProjectionType(*values)

Bases: Enum

Enum for different types of projections.

Attributes

RADEMACHER	Random binary projection matrix with values {-1, 1}.
GAUSSIAN	Random projection matrix with values drawn from N(0, 1/out_features).
ORTHOGONAL	Random orthogonal matrix with columns that form an orthonormal basis.
COMPLEX_GAUSSIAN	Complex-valued projection with real and imaginary parts from N(0, 1/(2*out_features)).
COMPLEX_ORTHOGONAL	Complex-valued orthogonal projection with orthonormal columns.

RADEMACHER = 'rademacher'

Random binary projection matrix with values {-1, 1}. Suitable for fast computation and memory-efficient implementations.

GAUSSIAN = 'gaussian'

Random projection matrix with values drawn from N(0, 1/out_features). Provides good theoretical guarantees for dimensionality reduction.

ORTHOGONAL = 'orthogonal'

Random orthogonal matrix with columns that form an orthonormal basis. Preserves angles and distances better than non-orthogonal projections.

COMPLEX_GAUSSIAN = 'complex_gaussian'

Complex-valued projection with real and imaginary parts from N(0, 1/(2*out_features)). Useful for complex signal processing and wireless communications.

COMPLEX_ORTHOGONAL = 'complex_orthogonal'

Complex-valued orthogonal projection with orthonormal columns. Provides optimal preservation of signal characteristics for complex data.