Releases: scikit-fuzzy/scikit-fuzzy
Scikit-Fuzzy version 0.4.2
This minor point release brings forward compatibility with NetworkX 2.4, which removed some prior methods of interactive with directed graphs. The changes remain backward compatible.
Thanks to @wouter-vdb for the fixes contained herein.
The release is live on PyPi, installable via pip
and will shortly be live on conda-forge.
Scikit-Fuzzy version 0.4.1
This minor point release brings a number of improvements, most importantly compatibility with Python 3.7. Also:
- The number of membership functions is no longer restricted in
automf
(#206) - Improvements to the documentation, especially for Consequents
In addition to the source release here, 0.4.1 is now live on PyPi.
Install/upgrade with
pip install -U scikit-fuzzy
Scikit-Fuzzy version 0.4.0
This major point release brings bug fixes, efficiency improvements, and quality of life improvements.
- A bug was fixed in the
lom
/som
methods; negative range now fully supported (#189) - Significant efficiency gains by streamlined linear algebra (#187) (#156)
- Names of nodes in the control system graph can now be displayed (#166)
- Numerical improvement to
cmeans
for values ofm
close to 1.0 (#154)
In addition to the source release here, 0.4.0 is now live on PyPi.
Install/upgrade with
pip install -U scikit-fuzzy
Scikit-Fuzzy 0.3.1
This minor point release brings a significant new feature and minor fixes/compatibility enhancements:
Major feature:
- Arrays are now accepted as system inputs; all inputs must have the same shape. The output matches this shape. This is dramatically more computationally efficient for repeated runs on existing data - within the limits of your system memory. (see #141)
Major fix
- Fixed the mathematical definition of
skfuzzy.gaussmf
; this is a potentially breaking change but the results are now correct (see #147).
Minor fixes
Scikit-Fuzzy 0.3
This major point release includes a number of improvements, primarily to the skfuzzy.control
submodule. A significant bug was squashed, the entire system is markedly more performant, and users can now more intuitively control the aggregation method.
- Fixed a memory leak on repeated, cached simulations not entirely flushing the cache (#120)
- AND/OR aggregation methods are now directly exposed to the user (#126)
np.interp
is used under the hood, resulting in major performance improvements (#130)- System visualization commands uniformly return both Matplotlib fig/axis objects, and better documented (#133/#136)
Thanks for using Scikit-Fuzzy, your reports keep making the package better!
Scikit-Fuzzy 0.2
This major point release brings many improvements to Scikit-Fuzzy, notably
- New fuzzy control system API
- Explore the new features located in
skfuzzy.control
. Recommended import statement isimport skfuzzy.control as ctrl
- note these new classes are not brought into the mainskfuzzy
namespace, they live only inskfuzzy.control
. - Designing complicated fuzzy control systems is now elegant, expressive, and Pythonic
- Peruse the new examples (reworked tipping problem and an advanced system) in the gallery
- Explore the new features located in
- New, more accurate defuzzification calculations result in higher accuracy with even sparser systems
- Significantly improved documentation and test coverage.