nearest_neighbors — Turi Create API 6.4.1 documentation (original) (raw)

The Turi Create nearest neighbors toolkit finds the rows in a tabular reference dataset that are most similar to a set of queries with the same schema.

A NearestNeighborsModel is created with a reference dataset contained in an SFrame, a distance function, and an indexing method (the latter two options can be done automatically by the model). An instantiated model has two key methods:query, for finding the closest points in the reference dataset to new data points; and similarity_graph, for finding the nearest neighbors of each point in the original reference set.

references = turicreate.SFrame({'x1': [0.98, 0.62, 0.11], ... 'x2': [0.69, 0.58, 0.36]}) references.print_rows() +------+------+ | x1 | x2 | +------+------+ | 0.98 | 0.69 | | 0.62 | 0.58 | | 0.11 | 0.36 | +------+------+ [3 rows x 2 columns] ... model = turicreate.nearest_neighbors.create(references) ... sim_graph = model.similarity_graph(k=1) sim_graph.edges +----------+----------+----------------+------+ | __src_id | __dst_id | distance | rank | +----------+----------+----------------+------+ | 0 | 1 | 0.376430604494 | 1 | | 2 | 1 | 0.55542776308 | 1 | | 1 | 0 | 0.376430604494 | 1 | +----------+----------+----------------+------+ ... queries = turicreate.SFrame({'x1': [0.05, 0.61, 0.99], ... 'x2': [0.06, 0.97, 0.86]}) queries.print_rows() +------+------+ | x1 | x2 | +------+------+ | 0.05 | 0.06 | | 0.61 | 0.97 | | 0.99 | 0.86 | +------+------+ [3 rows x 2 columns] ... model.query(queries, k=2) +-------------+-----------------+----------------+------+ | query_label | reference_label | distance | rank | +-------------+-----------------+----------------+------+ | 0 | 2 | 0.305941170816 | 1 | | 0 | 1 | 0.771556867638 | 2 | | 1 | 1 | 0.390128184063 | 1 | | 1 | 0 | 0.464004310325 | 2 | | 2 | 0 | 0.170293863659 | 1 | | 2 | 1 | 0.464004310325 | 2 | +-------------+-----------------+----------------+------+

In addition to the API documentation, please see the nearest neighbors chapter of the User Guidefor more details and extended examples.

nearest neighbors¶