- Interventions and counterfactuals out of the box to immediately allow you to generate actionable
recommendations to drive real outcomes.
- Trusted and transparent results enabling you to deliver with confidence.
- Choice and control for rapid customization to meet your business challenge.
CausalNet: State of the Art Structural Causal Modeling
Unlock interventions and counterfactuals to level up your decision making.
CausalNet is an advanced structural causal model (SCM) allowing you to push beyond the limitations of traditional machine learning algorithms. CausalNet gives you:
Explore new scenarios, and answer “what-if” questions immediately.
CausalNet provides you with simple, yet expressive, ways to work with interventions and counterfactuals.
- Interventions allow you to explore how changes in your data generating process would influence outcomes; supercharging your ability to plan for different scenarios.
- Meanwhile, counterfactuals provide retrospective analysis allowing you to drill into the “why” of a given instance.
- Decision Intelligence Engines bring interventions and counterfactuals together to immediately solve business challenges.
Deliver with confidence
Current AI relies on black box explainability methods such as SHAP or LIME. These methods only capture correlations learnt by the model, and rarely meet regulatory requirements.
Conversely, CausalNet allows you to:
- Maximize trust by inspecting all of the learned functions within CausalNet and see how changes to one set of variables influences others. Letting you observe how data flows through your system to identify problem areas.
- Minimize risk by constraining CausalNet with domain knowledge to guarantee known and predictable outcomes.
Customized to solve your business challenge.
CausalNet provides a breadth of functionality allowing you to configure your model to meet your own unique needs:
- Select from a range of advanced training engines such as PyTorch, CVXPY, DoubleML, or Pyro to meet your performance, data, and compute requirements while smoothing the route to production.
- Constrain training to match your domain expertise, by determining the types of functions which are applied to your edges and nodes.
- Handle tabular and time series data and learn both linear and non-linear relationships.