High-Performance Open-Source Archive
HnswEuclidean. This uses Euclidean distances
internally and will be returned from hnsw_build when
distance = "euclidean" is specified. This fixes an issue
where if you created an index with hnsw_build and
distance = "euclidean" (the default), then after saving,
you would be unable to reload the index and have it find Euclidean
distances. You would have to create it as an HsnwL2 object
and take the square root of the distances yourself (https://github.com/jlmelville/rcpphnsw/issues/21).Hnsw constructors and the hnsw_build
and hnsw_knn functions now expose a
random_seed parameter that you can use to set the random
seed used in constructing the HNSW index. If not provided, the the
hnswlib default of 100 is used (and should preserve
previous behavior). Based on a request by Maciej Beręsewicz (https://github.com/jlmelville/rcpphnsw/issues/23).addItemsCol,
getAllNNsCol and getAllNNsListCol are the
column-based equivalents of addItems,
getAllNNs and getAllNNsList, respectively.
Note that the returned nearest neighbor data from
getAllNNsCol and getAllNNsListCol are
also stored by column, i.e. the matrices have dimensions
k x n where k is the number of neighbors, and
n the number of items in the data being searched.byrow has
been added to hnsw_knn, hnsw_build and
hnsw_search. By default this is set to TRUE
and indicates that the items in the input matrix are found in each row.
To pass column-stored items, set byrow = FALSE. Any
matrices returned by hnsw_search and hnsw_knn
will now follow the convention provided by the value of
byrow: i.e. if byrow = FALSE, the matrices
contain nearest neighbor information in each column.getItems, which returns a matrix of the
data vectors in the index with the specified integer identifiers. From a
feature request made by d4tum (https://github.com/jlmelville/rcpphnsw/issues/18).progress parameter in the functional interface no
longer does anything. When verbose = TRUE, a progress bar
is no longer shown.setNumThreads method if using the object-based API, and the
n_threads parameter in the hnsw_* function
API. For finer control, a setGrainSize and
grain_size option is also available in the object and
function interface respectively. Thank you to Dmitriy Selivanov for a lot of
the work on this.verbose = TRUE now has incurs substantially
less computational overhead associated with calculating the progress
bar. Thank you to Samuel
Granjeaud for spotting the problem and coming up with various
solutions.progress. By default this is set to
"bar" and will show the progress bar when
verbose = TRUE. If you want a more terse output, set
progress = NULL. progress = NULL will
eventually be the default setting: for now, verbose = TRUE
will get you the progress bar by default for backwards
compatibility.markDeleted, that will remove an object
from being retrieved from the index.resizeIndex, that allows the index to be
increased without having to save and reload the index.size is available for the index objects
and reports the number of items added to the index.hnsw_search would stop if the number of
rows in the input matrix was smaller than k. This check has
been removed. Note that the correct behavior is to ensure that
k is smaller than or equal to index$size()
where index is the index you are searching. Because the
size() method is new to this version, to preserve
compatibility with old indexes, this check hasn’t been added to
hnsw_search. If this matters to you, manually compare
index$size() with k before running
hnsw_search. An error will be thrown if k
neighbors can’t be found in the index. Thank you to Yuxing Liao for spotting this and
the pull request to remove the check.Initial release.
Need mirroring services?
Contact our team at info@vpspulse.com.
Mirror powered by VPSpulse
Infrastructure sponsored by VPSPulse & Secure Payments by ArionPay.