protocol

ranker_protocol

Protocol for machine learning rankers.

Availability:
logtalk_load(ranking_protocols(loader))
Author: Paulo Moura
Version: 1:0:0
Date: 2026-05-06
Compilation flags:
static
Dependencies:
(none)
Remarks:
(none)
Inherited public predicates:
(none)

Public predicates

learn/3

Learns a ranker from the given dataset object using the specified options.

Compilation flags:
static
Template:
learn(Dataset,Ranker,Options)
Mode and number of proofs:
learn(+object_identifier,-compound,+list(compound)) - one

learn/2

Learns a ranker from the given dataset object using default options.

Compilation flags:
static
Template:
learn(Dataset,Ranker)
Mode and number of proofs:
learn(+object_identifier,-compound) - one

rank/3

Ranks a list of candidate items using the learned ranker and returns the candidates ordered from most preferred to least preferred.

Compilation flags:
static
Template:
rank(Ranker,Candidates,Ranking)
Mode and number of proofs:
rank(+compound,+list,-list) - one

scores/2

Returns the learned item-score pairs on the ranker native numeric scale. Depending on the ranker, these values may also be interpreted more specifically as strengths or ratings.

Compilation flags:
static
Template:
scores(Ranker,Scores)
Mode and number of proofs:
scores(+compound,-list(pair)) - one

diagnostics/2

Returns diagnostics and metadata associated with a learned ranker in a representation-independent way.

Compilation flags:
static
Template:
diagnostics(Ranker,Diagnostics)
Mode and number of proofs:
diagnostics(+compound,-list(compound)) - one

diagnostic/2

Tests or enumerates individual diagnostics metadata terms for a learned ranker.

Compilation flags:
static
Template:
diagnostic(Ranker,Diagnostic)
Mode and number of proofs:
diagnostic(+compound,?compound) - zero_or_more

ranker_options/2

Returns the effective training options recorded in a learned ranker diagnostics list.

Compilation flags:
static
Template:
ranker_options(Ranker,Options)
Mode and number of proofs:
ranker_options(+compound,-list(compound)) - zero_or_one

check_ranker/1

Checks that the argument is a structurally valid learned ranker term for the receiving ranker implementation, throwing an exception on invalid input when applicable.

Compilation flags:
static
Template:
check_ranker(Ranker)
Mode and number of proofs:
check_ranker(@compound) - one_or_error
Exceptions:
Ranker is a variable:
instantiation_error
Ranker is neither a variable nor a valid ranker:
domain_error(ranker,Ranker)

valid_ranker/1

True when check_ranker/1 succeeds for the argument without throwing an exception.

Compilation flags:
static
Template:
valid_ranker(Ranker)
Mode and number of proofs:
valid_ranker(@compound) - zero_or_one

export_to_clauses/4

Converts a ranker into a list of predicate clauses. Functor is the functor for the generated predicate clauses. When exporting a serialized ranker term, a noun such as ranker or model is usually clearer than a verb or result label.

Compilation flags:
static
Template:
export_to_clauses(Dataset,Ranker,Functor,Clauses)
Mode and number of proofs:
export_to_clauses(+object_identifier,+compound,+callable,-list(clause)) - one

export_to_file/4

Exports a ranker to a file. Functor is the functor for the generated predicate clauses. When exporting a serialized ranker term, a noun such as ranker or model is usually clearer than a verb or result label.

Compilation flags:
static
Template:
export_to_file(Dataset,Ranker,Functor,File)
Mode and number of proofs:
export_to_file(+object_identifier,+compound,+callable,+atom) - one

Protected predicates

(none)

Private predicates

(none)

Operators

(none)

See also

bradley_terry_ranker, ranking_dataset_protocol, pairwise_ranking_dataset_protocol