Real-Time Data-Driven Insights in Finance with OSS Hybrid AI
summary
Claims a hybrid OSS system that finds events by detecting simultaneous regime changes across many series using the Matrix Profile and a symbolic layer. Prototype (the Kronos project) runs 300k — with claims up to 10M — time series on a single server at subsecond rates, correlates hundreds of coincident changes to flag network or market events, learns causal precursors, and combines YOLO + openNARS for a working video collision/jaywalk detector. Open-source release planned Jan 2023.
AI-written summary · about
chapters (13)
- 0:00Real-time AI in finance
- 1:27Self-supervised time series
- 3:36Matrix profile & regime detection
- 7:30Interpreting regime changes
- 10:41Causal relationships & models
- 12:07Hierarchical knowledge models
- 14:55Symbolic reasoning & ontology
- 16:20Hybrid AI & cumulative learning
- 18:46Event detection demo
- 20:12Video monitoring with reasoning
- 21:34Unusual behavior detection
- 22:37OSS project & finance use cases
- 23:20Cisco Research team
related talks
13mBuild Real-Time Analytics Platforms for Financial Crime Detection
13mTransforming Financial Services with Real-Time Analytics
24mHunting Criminals with Hybrid Analytics
16mOperationalizing AI in Finance: Real-Time Analytics
52mDevil in the Haystack
36mDigital, data-driven financial advisory
11mUnlocking the full value of your data with operational analytics, ML & AI
13mBuilding Financial Fortresses: Real-Time Fraud Detection