RelationalSampler#

class sdm.relational.RelationalSampler(data: RelationalData, time_columns: Mapping[str, str] | None = None)#

Bases: object

Subgraph sampler over relational data.

Parameters:

  • data (RelationalData) – The collection of named tables and their relationships.

  • time_columns (Mapping[str, str] | None) – Mapping from table name to the datetime column used for temporal sampling. A row in a time-aware table can only be sampled if its timestamp does not exceed the query timestamp.

sample(task_table: TableTensor, task_link: TaskLink | Mapping[str, str | Sequence[str]], num_neighbors: Sequence[int], task_time_column: str | None = None, temporal_strategy: Literal['last', 'uniform'] = 'last') → RelationalSamplerOutput#

Sample RelatedTables for task rows.

Parameters:

  • num_neighbors (Sequence[int]) – Number of neighbors to sample per hop.

  • task_table (TableTensor) – Task table whose rows define the sampling queries.

  • task_link (TaskLink | Mapping[str, str | Sequence[str]]) – Link from task_table rows to a table in the relational data.

  • task_time_column (str | None) – Datetime column in task_table used as the query timestamp for temporal sampling.

  • temporal_strategy (Literal['last', 'uniform']) – How temporal neighbors are selected. "last" selects the most recent neighbors before each seed timestamp. "uniform" samples uniformly from neighbors before each seed timestamp.

Return type:

RelationalSamplerOutput