AWS, in a joint effort with Microsoft, has established PyWhy as a new GitHub organization to integrate AWS algorithms into DoWhy, a casual ML library from Microsoft, which has moved to PyWhy.
PyWhy’s mission is to build an open source ecosystem for causal machine learning that advances the state of the art and makes it available to professionals and researchers. PyWhy, will create and host libraries, tools, and other interoperable resources spanning a variety of causal tasks and applications, connected through a common API on fundamental causal operations and a focus on the end-to-end analysis process.
Most real-world systems, whether industrial procedures, supply chain systems, or distributed computer systems, can be characterized by variables that may or may not have a causal relationship to each other.
The evaluation of causal machine learning models and the formalization and integration of domain knowledge into machine learning pipelines present important research problems. Finding the best identification technique, creating an estimator, and performing robustness checks are phases that are often completed completely from scratch as part of the normal procedure. However, the hypotheses were difficult to understand and validate.
DoWhy is one of the existing causality libraries that focuses on various methods of estimating effects, with the overall goal of determining the impact of interventions on an objective variable.
Using the strength of graphical causal models, AWS work improves the GCMs of DoWhy’s current feature set. Judea Pearl, who won the Turing Prize, created the formal framework known as GCM for modeling causal links between variables in a system. Causal diagrams, which visually represent cause-effect links between variables observed with an arrow from a cause to their effect, are a crucial component of GCMs.
DoWhy already integrates possible causal graphical results and models, two of the most popular scientific frameworks for causal inference, into a single library for effect estimates. AWS’s contribution aims to strengthen the relationship between the frameworks and the communities of researchers involved.