0.8.0

matchms 0.8.0

Changed

• Minor changes to make tests pass with new matchms versions (>=0.18.0). It should nearly always be backward compatible though.
• Now dependency requirement is set to matchms>=0.14.0

0.7.0

Spec2Vec 0.7.0

Added

• added spec2vec.serialization subpackage to import and export Word2Vec models to/from disk without Pickle
  (via import_model and export_model respectively) #80

Changed

• bumped gensim version to >=4.2.0 in dependencies #84

Fixed

• updated Code examples in documentation to recent changes in matchms.

0.6.0

Spec2Vec 0.6.0

Main changes are the addition of proper logging and the changed handling of spectra with peaks (=words) missing in the model. Missing words will no longer raise exceptions but cause warnings and the return of empty embeddings if the percentage of missing words is above allowed_missing_percentage.

Added

• Logging (replacing former print statements) including options to write logs to file #73
• Now supports Python 3.9 (including CI test runs) #40

Changed

• missing words percentage above the allowed_missing_percentage no longer causes an expection but only leads to raising a warning #73
• default setting for allowed_missing_percentage to 10.0 to be less strict on model coverage #72

Fixed

• Can now also handle spectra in which no peak is known to the model (will return warning + empty vector) #73

0.5.0

This release is about moving from gensim 3.8 to gensim >= 4.0.0. This might give some performance improvements, but most of all it will be more future proof. It comes, however, with a few changes to the gensim API that might also influence some Spec2Vec scripts. In most cases things should work as before, but if issues arise, please first check the gensim information on how to migrate.

Changed

• Spec2Vec is now using gensim >= 4.0.0 #62

0.4.0

This releases expands the API and makes it possible to compute Spec2Vec similarities without the need to first create SpectrumDocument objects from Spectrum objects (will now be done as part of Spec2Vec).

Changed

• refactored Spec2Vec to now accept Spectrum or SpectrumDocument as input #51

Fixed

• updated and fixed code examples #51
• updated and fixed attribute typing #51

0.3.4

Mostly an attempt to solve an issue with the macOS conda workflow.

Changed

• update required numba version to >=0.51 to avoid issues between numba and numpy #55

0.3.3

Added

• Metadata getter method for SpectrumDocument #50
• Implement is_symmetric=True option for Spec2Vec.matrix method #53

Changed

• Change default for n_decimals parameter from 1 to 2 #50

0.3.2

Changed

• Add optional progress bar for spec2vec.matrix() calculations (default is False) #43

0.3.1

Changed

• Implement faster, numba-based cosine similarity function #29

0.3.0

Main changes for user

• use matchms.calculate_scores() instead of matchms.calculate_scores_parallel()
• use spec2vec.Spec2Vec instead of spec2vec.Spec2VecParallel
• use spec2vec.Spec2Vec.matrix(references, queries) instead of spec2vec.Spec2VecParallel(references, queries)

Added

• Support for Python 3.8 #35

Changed

• Refactored Spec2Vec class to provide .pair() and .matrix() methods #35

Removed

• Spec2VecParallel (is now included as Spec2Vec.matrix()) #35