December 2020 - Q4 has been a busy quarter for us at explainX. As we near the end of this year, we wanted to give our users something special. After feedback and constant iteration, we added four enhancement to our open-source that we believe will give the flexibility, power and customizability to enable every data scientist to ensure their machine learning models are trustworthy and successful in the real-world.
We will be releasing each enhancement every week as part of our open-source upgrades. This week, we are releasing explainX with a modular-structure to provide more flexibility and choice in utilizing the right explainability framework according to their AI use case. Each individual technique, be it shapley values or partial dependence or cohort analysis, will be accessible to the data scientists within their jupyter notebook. They will also have the option to retrieve data from these explainability techniques and integrate into their internal data applications like Tableua or PowerBI. Don't worry, we will be sharing tutorials on how to achieve this very soon!
We also learned that customizability and human-centered interfaces were extremely crucial for our users to truly understand the behavior of their machine learning models and share their insights with their business managers or clients. The end goal is to win the trust of the model consumer i.e. business analysts, business owners/product managers and deliver the maximum value from internal AI systems.
As our development thesis revolves around making it extremely simple and easier to explain blackbox machine learning models, we have stayed true to our 'single-line of code' access. Be it a neural network or a XGBoost, models can be explained with just a single line of code without any integration or customization issues. We want data scientists to focus on problem solving than spend their time on debugging their code and trying to make different frameworks work.
The first upgrade is live on GitHub as of today and we would like to invite all of you to head over, benefit from it, generate value and give us your feedback so we keep on helping you do your job 10x better!
Github Link: https://www.github.com/explainx/explainx
We will be posting tutorials and walkthroughs of explainX and how data scientists can get started in just a few minutes. Keep an eye open friends.