Rolling out DiCE for sklearn and regression models

• [Major] DiCE now supports sklearn models. Added three model-agnostic methods: randomized, genetic algorithm, and kd-tree
• [Major] Support for regression and multi-class problems
• [Major] Added local and global feature importance scores based on counterfactuals
• [Major] Better support for customizing counterfactuals through features_to_vary and permitted_range parameters for both continuous and categorical features
• [Refactor] ML Model and DiCE Explainer can use different feature transformations. Model's transformation can be provided as an input to the dice_ml.Model constructor. DiCE accepts inputs in the original data frame and does its transformations internally
• Enhanced tests for the library
• Deep learning libraries (tensorflow and pytorch) marked as optional dependencies
• New notebooks showing applications of DiCE in docs/source/notebooks/

A big thanks to @raam93, @soundarya98 and @gaugup for this release!

v0.4 : Faster VAE-based method for CFs, Pytorch and Tensorflow 2.x support

Here's the latest stable version.

• DiCE now supports Pytorch, Tensorflow 1.x and 2.
• Includes a Variational AutoEncoder-based method to generate counterfactual examples, based on https://arxiv.org/abs/1912.03277. This method is much faster--try it out!
• Support for private data, when only aggregate training data statistics are available to generate counterfactuals
• Updated and faster post-hoc sparsity enhancer module for counterfactuals
• Includes bug fixes for DiCE and tests for most DiCE functionalities.
• More notebooks and detailed docs at http://interpret.ml/DiCE/

Big thanks to @raam93 for leading the updates, and to @divyat09 for adding the VAE method.

First release

Supports counterfactual explanations for tensorflow and pytorch classifiers.