Release Notes¶

NetworkUnit 0.2.0¶

parameter handling
- generate_prediction() and other custom class function no longer take optional extra parameter as arguments, but only use self.params
- no class function should accept arguments that override class parameters
- default_params test class attribute are inherited by using default_params = {–parent.default_params, ‘new_param’:0}
caching
- improved caching of intermediate test- and simulation results, e.g. for the correlation matrix
- improving backend definitions
parallelization
- automatic parallelization for loops over spiketrains or lists of spiketrains. To use set params[‘parallel executor’] to ProcessPoolExecutor(), MPIPoolExecutor(), or MPICommExecutor() (see [documentation in Elephant package](https://elephant.readthedocs.io/en/latest/reference/parallel.html))
various bug fixes
new features
- adding the joint_test class that enables the combination of multiple neuron-wise tests for multidimensional testing with the Wasserstein score
new test classes
- joint_test
- power_spectrum_test
  
  freqband_power_test
- timescale_test
- avg_std_correlation_test
new score classes
- wasserstein_distance
- eigenangle (see publication [Gutzen et al. 2022](https://doi.org/10.1016/j.biosystems.2022.104813)

a fix for an issue where the setup script was failing to properly install the backend directory (see issue #20)

a new backend class, which handles the storage of generated predictions in memory or on disk. To make use of it just set backend=’storage’ in the model instantiation. By default predictions are stored in memory. To change that set `model.get_backend().use_disk_cache = True ` and `model.get_backend().use_memory_cache = False `.
various bug fixes
updated requirements.txt and environment.yaml

Initial release.