Release History¶
Version 0.5.1¶
First public release.
Version 0.5.0¶
Features¶
TPE optimization with Optuna
Version 0.4.1¶
Features¶
Features filtering can be done by variance, Median Absolute Deviation (MAD) or truncated Single Value Decomposition (t-SVD)
Integrated UMAP dimensionality reduction
Supervised dimensionality reduction with UMAP
DBSCAN, HDBSCAN, Shared Nearest Neighbours and Louvain community detection are available for clusters identification
k-Nearest Neighbours classification
Parameters range selection can be automatic, for large jobs the mesh can be dynamically adapted
Old runs can be resumed or extended for a finer search
New data can be added to update an old hierarchy of clusters
A short tutorial with the MNIST dataset is available
GPU implementation with RAPIDS