Personalized

If there are no signals in the current session, the personalized batch functions as a random batch.

Personalized batches serve personalized data based on session signals. They use N signals to calculate M vectors and M weights.

N is the number of signals created by user actions.

M is the number of vectors created by the Personalized Batch operation. M depends on a few things including:

• How signals are positioned in the embeddings space

• Parameters

These M vectors are queried to vector databases to fetch batch_size results.

Personalized Batch

Accepted parameters for Personalized batches are:

mu: float = 0.0
alpha: float = 0.0
r: float = 0.0
last_n: int = 0
remove_duplicates: bool = True
apply_threshold: Tuple[bool, float] = (False, 0.0)
apply_mmr: bool = False