…n layer weights for each feature as it evolves

over epochs, to see effect of setting weight decay parameters; (2) Added model parameters xgb_alpha and xgb_lambda
and corresponding hyperopt parameters xgba and xgbb to control strength of L1 and L2 regularization penalties.

…hing on weight_decay_penalty and

weight_decay_penalty_type parameters. Completed implementation of hyperopt domain specification for
xgb_alpha and xgb_lambda.

…penalty_type, xgb_alpha and xgb_lambda to set

of model parameters displayed by the various compare_models functions.

… step(); changed order of operations so that grading

happens at end of step() method rather than at beginning. Added serial_grade_population method for debugging. Simplified
code to address some performance issues.

…eeing that it runs much faster than

the multithreaded version. Added documentation.

…rity between test and training

set scaffold structures. Fixed a bug where the splitter always returned the split from the last
generation rather than the best-ever split. Added code to track the individual fitness function terms
over generations so that they can be displayed in diagnostic plots.

…ore to the fitness_scores dictionary so that it can

be plotted together with the component scores.

…specified number of generations. Replaced print() calls

with log messages so we can control verbosity of output. Changed split() to use log_every_n argument to control frequency
of messages during GA operation.

…nd XGBoost classification models now support

class balancing weights when weight_transform_type parameter is set to 'balancing'.

… valid, test.

…ght absolute sums vs epoch,

to assess effect of weight decay penalty.

…n layer weights for each feature as it evolves

over epochs, to see effect of setting weight decay parameters; (2) Added model parameters xgb_alpha and xgb_lambda
and corresponding hyperopt parameters xgba and xgbb to control strength of L1 and L2 regularization penalties.

…hing on weight_decay_penalty and

weight_decay_penalty_type parameters. Completed implementation of hyperopt domain specification for
xgb_alpha and xgb_lambda.

…penalty_type, xgb_alpha and xgb_lambda to set

of model parameters displayed by the various compare_models functions.

… step(); changed order of operations so that grading

happens at end of step() method rather than at beginning. Added serial_grade_population method for debugging. Simplified
code to address some performance issues.

…eeing that it runs much faster than

the multithreaded version. Added documentation.

…nd XGBoost classification models now support

class balancing weights when weight_transform_type parameter is set to 'balancing'.

… valid, test.

…ght absolute sums vs epoch,

to assess effect of weight decay penalty.