recency_proportional_resolution_curbed_algo
Eagerly utilize fixed stratum storage capacity size_curb to minimize recency-proportional MRCA uncertainty.
The curbed recency-proportional resolution policy ensures that retained stratum count respects a user-specified size cap. Recency-proportional MRCA resolution is provided at finest-possible granularity given space constraint with graceful degradation of granularity occuring as deposition history grows.
This policy works by splicing together successively-sparser recency_proportional_resolution_algo parameterizations then — after the stratum retention count upper bound of the sparsest recency_proportional_resolution_algo exceeds storage capacity — a permanently fixed parameterization of the geometric_seq_nth root algorithm.
While
num_depositions < 2 ** (size_curb // 3) // 2,
recency_proportional_resolution_algo policies apply with resolution
size_curb // ceil(log2(num_depisitons)) - 1.
For
num_depisitons > 2 ** (size_curb // 3) // 2,
a geometric_sequence_nth_root_algorithm policy with degree
(size_curb - 2) // 6
and interspersal 2 applies. To ensure availability of ranks required by the geometric_sequence_nth_root_algorithm, the transition between algorithms occurs exactly before recency-proportional resolution 1 would apply.
Because strata retained by each policy supersets strata retained by all subsequent retention policies, uncertainty bounds for all policies apply within their respective domains. (I.e., all strata expected by each policy are available when that policy activates.)
Although the geometric sequence nth root algorithm also guarantees constant space complexity with recency-proportional MRCA uncertainty, the curbed recency- proportional algorithm makes fuller (i.e., more aggressive) use of available space during early depositions.
The curbed recency-proportional algorithm also compares favorably to the tapered geometric sequence nth root algorithm during early depositions. Although the tapered nth root algorithm makes the fullest (i.e., most aggressive) use of available space — upper bound capacity is always perfectly utilized — the used space is less effective at minimizing worst-case recency- proportional resolution than the curbed algorithm. (I.e., the tapered nth root algorithm provides patchier, less systematic coverage.)
Notes
Only size caps of eight strata and above are guaranteed respected. (Initialization of smaller size caps will warn.)
See Also
- recency_proportional_resolution_algo:
Provides log space complexity and recency-proportional strata spacing.
- geom_seq_nth_root_algo:
Provides constant space complexity and MRCA rank estimate uncertainty bounded under s * (1 - n^(-1/k)) where n is the number of strata deposited and s is the true number of ranks deposited since the MRCA.
- geom_seq_nth_root_tapered_algo:
Perfectly space-filling variant of geom_seq_nth_root_algo.
Classes
Functor to provide member function implementation in Policy class. |
|
alias of |
|
Functor to provide member function implementation in Policy class. |
|
alias of |
|
alias of |
|
Functor to provide member function implementation in Policy class. |
|
Functor to provide member function implementation in Policy class. |
|
Functor to provide member function implementation in Policy class. |
|
Functor to provide member function implementation in Policy class. |
|
Functor to implement the recency-proportional resolution stratum retention policy, for use with HereditaryStratigraphicColumn. |
|
Functor to provide member function implementation in Policy class. |
|
alias of |
|
Contains all policy parameters, if any. |
- class CalcMrcaUncertaintyAbsExact
Functor to provide member function implementation in Policy class.
- __call__(policy: PolicyCouplerBase, first_num_strata_deposited: int, second_num_strata_deposited: int, actual_rank_of_mrca: int) int [source]
Exactly how much uncertainty to estimate rank of MRCA?
- __init__(policy_spec: PolicySpec | None) None [source]
- CalcMrcaUncertaintyAbsUpperBound
alias of
CalcMrcaUncertaintyAbsUpperBoundWorstCase
- class CalcMrcaUncertaintyAbsUpperBoundAtPessimalRank
- class CalcMrcaUncertaintyAbsUpperBoundPessimalRank
Functor to provide member function implementation in Policy class.
- __call__(policy: PolicyCouplerBase, first_num_strata_deposited: int | None, second_num_strata_deposited: int | None) int [source]
Calculate rank for which upper bound on absolute MRCA uncertainty is pessimized.
- __init__(policy_spec: PolicySpec | None) None [source]
- CalcMrcaUncertaintyRelExact
alias of
CalcMrcaUncertaintyRelExactFromAbs
- CalcMrcaUncertaintyRelUpperBound
alias of
CalcMrcaUncertaintyRelUpperBoundWorstCase
- class CalcMrcaUncertaintyRelUpperBoundAtPessimalRank
- class CalcMrcaUncertaintyRelUpperBoundPessimalRank
Functor to provide member function implementation in Policy class.
- __call__(policy: PolicyCouplerBase, first_num_strata_deposited: int, second_num_strata_deposited: int) int [source]
Calculate rank for which upper bound on relative MRCA uncertainty is pessimized.
- __init__(policy_spec: PolicySpec | None) None [source]
- class CalcNumStrataRetainedExact
Functor to provide member function implementation in Policy class.
- __call__(policy: PolicyCouplerBase, num_strata_deposited: int) int [source]
Exactly how many strata are retained after n deposited?
- __init__(policy_spec: PolicySpec | None) None [source]
- class CalcNumStrataRetainedUpperBound
Functor to provide member function implementation in Policy class.
- __call__(policy: PolicyCouplerBase, num_strata_deposited: int) int [source]
At most, how many strata are retained after n deposited? Inclusive.
- __init__(policy_spec: PolicySpec | None = None) None [source]
- class CalcRankAtColumnIndex
Functor to provide member function implementation in Policy class.
- __call__(policy: PolicyCouplerBase, index: int, num_strata_deposited: int | None) int [source]
After n strata have been deposited, what will the rank of the stratum at column index k be?
Enables a HereditaryStratigraphicColumn using this predicate to optimize away storage of rank annotations on strata. Takes into the account the possibility for in-progress stratum depositions that haven’t been reflected in num_strata_deposited.
- __init__(policy_spec: PolicySpec | None) None [source]
- class GenDropRanks
Functor to implement the recency-proportional resolution stratum retention policy, for use with HereditaryStratigraphicColumn.
This functor enacts the recency-proportional resolution policy by specifying the set of strata ranks that should be purged from a hereditary stratigraphic column when the nth stratum is deposited.
- __call__(policy: PolicyCouplerBase, num_stratum_depositions_completed: int, retained_ranks: Iterable[int] | None) Iterator[int] [source]
Decide which strata within the stratagraphic column should be purged.
Every time a new stratum is deposited, this method is called to determine which strata should be purged. All strata at ranks yielded from this functor are immediately purged from the column, meaning that for a stratum to persist it must not be yielded by this functor each and every time a new stratum is deposited.
Parameters
- policy: Policy
Policy this functor enacts.
- num_stratum_depositions_completedint
The number of strata that have already been deposited, not including the latest stratum being deposited which prompted the current purge operation.
- retained_ranksiterator over int
An iterator over ranks of strata currently retained within the hereditary stratigraphic column. Not used in this functor.
Returns
- iterator over int
The ranks of strata that should be purged from the hereditary stratigraphic column at this deposition step.
See Also
- recency_proportional_resolution_curbed_algo:
For details on the rationale, implementation, and guarantees of the recency-proportional resolution stratum retention policy.
- __init__(policy_spec: PolicySpec | None) None [source]
- class IterRetainedRanks
Functor to provide member function implementation in Policy class.
- __call__(policy: PolicyCouplerBase, num_strata_deposited: int) Iterator[int] [source]
Iterate over retained strata ranks at num_strata_deposited in ascending order.
- __init__(policy_spec: PolicySpec | None) None [source]
- class PolicySpec
Contains all policy parameters, if any.