Hepynet

Ver 0.4.0 updates:
* Use a new name for the project
* Using logging as the log printer
* Move root/NumPy transfer function to an individual package
* Use pathlib to handle path
* Refactor
* Optimize memory use
* Test configurations and clean-up codes
* Update README for the new release

For the next releases:
* Flexible array size, (for example, you can specify use 10% of inputs for training or analyzing)
* Customized model class
* Unify NumPy array to use 2D array (don't allow (n, ) shape, use (n, 1) instead )*
* Set up formatter*
* Prepare the Wik

Ver 0.3.0 updates

Refactor for new ntuples:
* Updates data IO
* Updates npy storage
* Updates validation functions
** Apply trained model to test
** Apply trained model to data

Add new functions:
* Auto scan evaluation:
** Automatically summary scan result
** Find the best selection
* Add k-fold validation utilities （abandoned）
* Add more automatic validation functions:
** Kinematic plots (use HEPTool package)
** Add DNN score plot function (use HEPTool)
** Add ratio plot for data/bkg comparison (use HEPTool)
** Correlation matrix
** Significance
* Add model with simpler structure:
** 3 layers, 100 nodes per layer
* Add model with dropout layer

Final refactor:
* Update code style: naming, comment style, etc.
** Move ratio plot function to th1_tools
** Move other general plot functions to plot_utils
** Use initialize dictionary to initialize model to deal with different inputs in different models
* Update documentations
* Update the example

Test:
*Test framework v0.3.0 in Z prime analysis

Major new features:
* Job executer: Reads the configuration file and execute jobs automatically. Can perform parameter-scan to find the best training parameters.
* Auto-report: Automatically generate training reports to show machine details and training performance. More diagnostic plots added.