Shuffling for Input Parameter Dimensionality Reduction

• Added option to shuffle data sets for each population member when limiting input parameter dimensionality using genetic algorithm
• Removed shuffling functionality for iterative inclusion dimensionality reduction; caused indexing errors

Bug fixes for 1.4.1

Moved DataPoint class out of DataFrame class in ecnet/data_utils.py; having this inside caused errors in saving DataPoint objects using save_project() Server method

Bug fixes and exception handling for 1.4.0

• Removed extraneous lines in error_utils.py
• Added exceptions for empty sets in model.py

Overhaul of ECNet source files

• Update to server.py syntax (some method names have been changed, compare your scripts to the new examples!)
• Overhauled data_utils.py
• Reformatted model.py
• Inclusion of genetic algorithm for input feature dimensionality reduction
• Artificial bee colony functionality for hyperparameter tuning now relies on ECabc Python package, it is no longer integrated into ECNet
• Update to database format (your old databases will need to be edited to remain compatible!)

Integration of Hyperparameter Tuning

Neural network training hyperparameters can now be tuned using an artificial bee colony optimization algorithm. For more on artificial bee colonies: https://en.wikipedia.org/wiki/Artificial_bee_colony_algorithm

Limit parameter bug fix

Fixed issues related to limiting the input parameter dimensionality.
Files edited:

• ecnet/limit_parameters.py
• ecnet/server.py
• setup.py