DeepCausality still has some more work to do because of its early stage. In its current state, a handful of limitations exist:
Counterfactual reasoning is missing. DeepCausality cannot reason counter to the fact; for example, if the drone had not accelerated more than 50mp/h, it would not have crashed into the tree.
Causal structural learning is missing. Right now, causal models have to be designed and built by hand. Possible areas of exploration for causal structural learning are:
The author of DeepCausality has a background in Fintech and spent considerable time modeling market volatility. Market volatility modeling can be done in one of two ways. Mathematical models are still considered the gold standard, and many institutions operate them in production. Because markets work on time-series data, quantitative analysts (quants) use many time-series statistical techniques (ARIMA) combined with various differential equations. While these models are complex and require substantial optimization and computational power, they are robust and reliable in operations.
Overview Four layers of causal reasoning Observations Assumptions Inference Causality Core Architecture Context architecture Implementation principles About Overview DeepCausality follows a first principal architecture combined with a handful implementation principles to achieve a high degree of functionality with efficient and straightforward implementation. DeepCausality’s architecture uses the four layers of causal reasoning to achieve these goals.
Four layers of causal reasoning Conceptually, causal reasoning relies on four pillars: observations, assumptions, inferences, and causes, as depicted in the diagram below.
Overview Context Relations between data Conceptualization of time, space, and spacetime Data Time Space SpaceTime Structural conceptualization of causation Transparent composability Non-Euclidian data representation About Overview DeepCausality is a hyper-geometric computational causality library that enables fast and deterministic context-aware causal reasoning. Unique to DeepCausality is its ability to contextualize causal models, allowing causal reasoning over data from multiple external sources aggregated into one context. Beyond that, DeepCausality contributes several novel concepts to computational causality:
Overview Neuronal Nets The universal approximation theorem Independent and identically distributed data (IID) assumption Beyond the IID The gap Lack of context Lack of relations between data Lack of conceptualization of time and space Lack of structural conceptualization of causation Lack of non-Euclidian data representation Lack of transparent composability About Overview With the advent of advanced large language models, many new opportunities arise for applied artificial intelligence, but some challenges are becoming apparent.
Overview What is computational causality? Why Rust? What is DeepCausality? Hypergeometry Context Contextual causal reasoning End-to-end explainability Causal State Machines What can you do with DeepCausality? Contextualized streaming data Financial modelling Dynamic control systems Combined deep causality learning About Overview DeepCausality is a hyper-geometric computational causality library that enables fast and deterministic context aware causal reasoning. Deep causality offers multiple benefits such as fast reasoning, low-cost operations due to low computational requirements, the ability to analyse numerous data feeds in real-time, and even reasoning across data that changes over time, space, and spacetime through an adjustable context.
