Trait sbr::OnlineRankingModel

pub trait OnlineRankingModel {
    type UserRepresentation: Debug;
    fn user_representation(
        &self, 
        item_ids: &[ItemId]
    ) -> Result<Self::UserRepresentation, PredictionError>;
    fn predict(
        &self, 
        user: &Self::UserRepresentation, 
        item_ids: &[ItemId]
    ) -> Result<Vec<f32>, PredictionError>;
}

Trait describing models that can compute predictions given a user's sequences of past interactions.

Associated Types

type UserRepresentation: Debug

The representation the model computes from past interactions.

Required Methods

fn user_representation(
    &self,
    item_ids: &[ItemId]
) -> Result<Self::UserRepresentation, PredictionError>

Compute a user representation from past interactions.

fn predict(
    &self,
    user: &Self::UserRepresentation,
    item_ids: &[ItemId]
) -> Result<Vec<f32>, PredictionError>

Given a user representation, rank item_ids according to how likely the user is to interact with them in the future.

Implementors

• impl OnlineRankingModel for ImplicitEWMAModel type UserRepresentation = ImplicitUser;

• impl OnlineRankingModel for ImplicitLSTMModel type UserRepresentation = ImplicitUser;

• impl<U: SequenceModel, T: SequenceModelParameters<Output = U> + Sync> OnlineRankingModel for T type UserRepresentation = ImplicitUser;