Bletl

Latest version: v1.5.1

Safety actively analyzes 682361 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 4 of 4

1.0.3

This release brings bugfixes and a tiny breaking change related to the `bletl.growth.fit_mu_t` function:
+ The model was refactored to describe **µ only for the segments inbetween** the data points. This removes a bias introduced by incorrectly shifting the prediction by 1 timestep. Unfortunately this also means that **X and mu have different lengths**, so the `GrowthRateResult.t` **was removed** in favor of `GrowthRateResult.t_data` and `GrowthRateResult.t_segments`.
+ A `mu_prior` kwarg was added to `fit_mu_t`. This allows the user to predefine where the random walk should start and can greatly improve the result on datasets that don't start with a lag phase.
+ The `σ` kwarg was renamed to `drift_scale` and is no longer optional.
+ The probability threshold used for switchpoint detection can now be user-configured.
+ Two bugs in filtering detected switchpoints were fixed.
+ Initial growh rate guesses are now clipped to the interval [-0.5, 0.5].

See https://github.com/JuBiotech/bletl/pull/4 and https://github.com/JuBiotech/bletl/pull/5 for more information.

1.0.2

+ Fixes a bug in the `bletl.growth._make_random_walk` helper function that caused the `nu` kwarg to be ignored.
+ Also adds a `nu` kwarg to the `fit_mu_t` function to allow for fine-tuning of switchpoint prior probability.

Also see 29bf6db4c0226810072dd1f7b22c18ecda1a5c21.

1.0.1

+ Fixes a bug where `fit_mu_t` without predefined switchpoints ran into an error
+ Increases robustness of `fit_mu_t` initialization with non-linear calibration models.

See 1fd14a0e694971472007de59917dd6a28eef98ee for details.

1.0.0

This is the first public release.

Changes w.r.t. previous internal releases:
+ **`bletl_analysis` was merged into the `bletl` package.**
+ Sphinx-based documentation was added (see https://bletl.readthedocs.io).
+ `fit_mu_t` now initializes with a growth rate vector calculated by a smoothing approach. This results in more robust optimization.
+ Unit tests were refactored to use `pytest`.
+ Docstring were converted to NumpyDoc style.
+ Example notebooks were updated.
+ The `bletl.core` submodule was refactored into `bletl.types` and logic from `bletl/__init__.py` was moved to `bletl.core`.

Page 4 of 4

© 2024 Safety CLI Cybersecurity Inc. All Rights Reserved.