Changelogs » Pyndl



* fixes math.ceil with explicit cast to double before ceiling

* makes error codes in ndl_parallel more human readable


* fixes error code -166... (issue 85 on github)
* the (now fixed) error lead to pyndl not running under specific circumstances
* general code improvements


- fixes encoding error in long description of
- a lot of small code improvements
- improvements in the maintenance code


* fixes integer overflow in uni-dimensional index of weights in ``ndl.ndl`` that results in wrong learning in weight matrices with more than 4294967295 cells
* please relearn any weights where number of cues times numbers of outcomes exceeds 4294967295 (sorry for this bug)
* thanks to Yuying for finding it


- Moves code to read and write events into ```` module. This changes the existing API! Deprecation warnings are added accordingly.
- improves documentation


* removes duplicate outcomes, when remove_duplicates=True was given in (thanks to Christian Adam kuchenrolle for spotting and fixing the problem)


- fixes bug in pyndl.ndl which did not run the events in order
- fixes pylint errors and improves code quality
- excludes files from zip and tar.gz archives


* fixes memory leak in pyndl.ndl.ndl
* fixes problem with pyndl.sysinfo()
* enhances pyndl.ndl.ndl by allowing to specify the temporary directory (thanks to Christian Adam)


fixes `` to make package pip installable.


Gets rid of older code and removes some old API. Improves documentation a lot. In more detail:

* removes incomplete code for Kalman filtering
* removes command line interface for
* heavily improves documentation
* adds citation information
* cleanups of configuration files
* does not depend on matplotlib and docopt anymore


New function `pyndl.sysinfo()` is added to provide details in case of bug reports.


Pyndl does no longer support Python 3.4 since pandas does not support it anymore. We recommend upgrading your Python installation to at least 3.5!


fixes zenodo meta data


Minor release which mainly fixes maintaining issues.

* use `pycodestyle` instead of `pep8`
* add meta information for [zenodo](


- gzipped event files (solves 68)
- harmonize dict_ndl and ndl (solves 75)
- replace docopt with dependecy (solves 71)
- improve readme (solves 80)
- improve alpha attribute in dict_ndl (solves 73)

If you want to reuse old event files please gzip them in order to use them with the new code.

The following list contains the API changes in the format old function call above and new function call below:

activation.activation(event_list, ...)
=> activation.activatoin(events, ...)
=> ndl.events_from_file(...)

ndl.ndl(event_path, ...)
=> ndl.ndl(events, ..., verbose=False)

ndl.dict_ndl(event_list, ...)
=> ndl.dict_ndl(events, ..., verbose=False)

ndl.slice_list(..., sequence)
=> ndl.slice_list(..., len_sublists)


- makes activation function consistent with the output of ndl and dict_ndl
- dict_ndl no longer produces NaNs for cue-outcome combinations which did not occur if `make_data_array=True`
- dimensions of the output matrix of activation was changed to outcomes, events


* fixes bug with missing cues in unseen events
* deletes temporary files after use


improves meta data and fixes dict_ndl for specific inputs


removes claimed support for python 3.2 and 3.3, which was never tested