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.
The second way emerged with the advent of artificial intelligence. The idea is relatively simple: train an AI model on market data and successful trades, then the AI supposedly trades successfully. In practice, this rarely works, although the reasons are rooted mainly in how contemporary AI works internally. There are several impediments that make conventional artificial intelligence unsuitable for financial markets. The background section details how these three impediments affect modern machine learning. Solving one of them is challenging but possible. However, resolving all of them would require rethinking the entire foundation of machine intelligence.
A lesser-known alternative to conventional artificial intelligence is the field of computational causality, which has its foundation in cause-effect relations. However, while computational causality has existed for some time, it is dominated by academic research and few industry research groups, mainly because no clear-cut methodology has yet emerged. As a response to the lack of production-ready libraries, the author invented DeepCausality to build what was missing in Rust.
The core concept of DeepCausality is rooted in advanced theoretical physics in which causal models are used to establish a unified structure for Quantum Gravity. Lucian Hardy at the Perimeter Institute singlehandedly pioneered and published a novel conceptualization of causality (summary) which serves as a foundation in the quest to formalize quantum gravity. Without this groundbreaking contribution, there would not have been a conceivable way of making advanced computational causality operational for comprehensive industry applications.
However, it took a few more innovations to make DeepCausality a reality. Among them are context, hyper geometric causal structures, and contextual causal reasoning. For details, see the introduction, architecture, and concept sections of the documentation.
Before DeepCausality, complex market models were challenging to build, hard to understand, and required a lot of expensive hardware. With DeepCausality the process is easier and faster by orders of magnitude. Capturing intricate patterns in up to four-dimensional spaces and analyzing them in relation to multiple fast changing contexts actually becomes feasible. Once the first version of DeepCausality was completed, it was clear that its application reaches well beyond the financial industry and, in response, the author donated DeepCausality to the Linux Foundation so all of mankind could benefit from this new technology.