Pipeline

Pipeline

Pipeline(*, trust_propagation: TrustPropagation = trust_propagation, preference_learning: PreferenceLearning = preference_learning, voting_rights: VotingRightsAssignment = voting_rights, scaling: Scaling = scaling, aggregation: Aggregation = aggregation, post_process: PostProcess = post_process)

Instantiates the pipeline components.

Parameters:

Name	Type	Description	Default
trust_propagation	TrustPropagation	Algorithm to spread trust based on pretrusts and vouches	trust_propagation
preference_learning	PreferenceLearning	Algorithm to learn a user model based on each user's data	preference_learning
voting_rights	VotingRightsAssignment	Algorithm to assign voting rights to each user	voting_rights
scaling	Scaling	Algorithm to put user models on a common scale	scaling
aggregation	Aggregation	Algorithm to aggregate the different users' models	aggregation
post_process	PostProcess	Algorithm to post-process user and global models, and make it readily usable for applications.	post_process

run

run(input: PipelineInput, criterion: str, output: Optional[PipelineOutput] = None)

Executes the pipeline with the given input and criterion.

Parameters:

Name	Type	Description	Default
input	PipelineInput	The input data for the pipeline.	required
criterion	str	The criterion used for the pipeline execution.	required
output	Optional[PipelineOutput]	The output object to store results, by default None	None

__call__

__call__(users: DataFrame, vouches: DataFrame, entities: DataFrame, privacy: PrivacySettings, judgments: Judgments, init_user_models: Optional[dict[int, ScoringModel]] = None, output: Optional[PipelineOutput] = None) -> tuple[DataFrame, VotingRights, Mapping[int, ScoringModel], ScoringModel]

Run Pipeline

Parameters:

Name	Type	Description	Default
users	DataFrame	user_id: int (index) is_pretrusted: bool	required
vouches	DataFrame	voucher (int) vouchee (int) vouch (float)	required
entities	DataFrame	entity_id: int (index)	required
privacy	PrivacySettings	privacy[user, entity] in { True, False, None }	required
judgments	Judgments	judgments[user] must yield the judgment data provided by the user	required
init_user_models	Optional[dict[int, ScoringModel]]	user_models[user] is the user's model	None

Returns:

Name	Type	Description
users	DataFrame with columns	user_id: int (index) is_pretrusted: bool trust_score: float
voting_rights	VotingRights	voting_rights[user, entity] is the user's voting right for entity
user_models	dict[int, ScoringModel]	user_models[user] is the user's model
global_model	ScoringModel	global model

DefaultPipeline

DefaultPipeline

Instantiates the default pipeline described in "Solidago: A Modular Pipeline for Collaborative Scaling".

trust_propagation

trust_propagation: TrustPropagation = LipschiTrust(pretrust_value=0.8, decay=0.8, sink_vouch=5.0, error=1e-08)

preference_learning

preference_learning: PreferenceLearning = UniformGBT(prior_std_dev=7, convergence_error=1e-05, cumulant_generating_function_error=1e-05)

voting_rights

voting_rights: VotingRightsAssignment = AffineOvertrust(privacy_penalty=0.5, min_overtrust=2.0, overtrust_ratio=0.1)

scaling

scaling: Scaling = ScalingCompose(Mehestan(lipschitz=0.1, min_activity=10.0, n_scalers_max=100, privacy_penalty=0.5, p_norm_for_multiplicative_resilience=4.0, error=1e-05), QuantileZeroShift(zero_quantile=0.15, lipschitz=0.1, error=1e-05), Standardize(dev_quantile=0.9, lipschitz=0.1, error=1e-05))

aggregation

aggregation: Aggregation = EntitywiseQrQuantile(quantile=0.2, lipschitz=0.1, error=1e-05)

post_process

post_process: PostProcess = Squash(score_max=100)