|
Uni20
Asynchronous tensor and runtime library.
|
Rank-selection policy for a truncated singular value decomposition. More...
#include <truncated_svd.hpp>
Public Attributes | |
| std::optional< Real > | maximum_discarded_weight = std::nullopt |
| std::size_t | maximum_retained_extent = uni20::numeric_limits<std::size_t>::max() |
| std::size_t | minimum_retained_extent = 0 |
| std::optional< Real > | normalized_squared_singular_value_cutoff = std::nullopt |
| std::optional< Real > | singular_value_cutoff = std::nullopt |
Rank-selection policy for a truncated singular value decomposition.
Active cutoff and discarded-weight criteria each impose a minimum rank. Their maximum is then clamped by the explicit extent bounds. The maximum retained extent is a hard cap and may therefore prevent another requested criterion from being satisfied.
| std::optional<Real> uni20::linalg::SvdTruncationPolicy< Real >::maximum_discarded_weight = std::nullopt |
| std::size_t uni20::linalg::SvdTruncationPolicy< Real >::maximum_retained_extent = uni20::numeric_limits<std::size_t>::max() |
| std::size_t uni20::linalg::SvdTruncationPolicy< Real >::minimum_retained_extent = 0 |
| std::optional<Real> uni20::linalg::SvdTruncationPolicy< Real >::normalized_squared_singular_value_cutoff = std::nullopt |
| std::optional<Real> uni20::linalg::SvdTruncationPolicy< Real >::singular_value_cutoff = std::nullopt |
1.9.8