Machine Learning
Stream Learning for Quants & Traders
A mathematics-first architecture which enables continuous "in-flight" model optimisation
HISTORY
Disc is the next step in the evolutionary chain of trading technology
Stream Learning
Disc Connects Cause to Effect
By creating a positive feedback loop, Disc architecture removes all stages points between abstract theory and real world trading.
The entire Disc architecture and infrastructure focuses on creating end-to-end streamed data lineage throughout the entire tech stack
Using this architecture, Disc creates self-correcting models which predict, test and refine - all in real time, using a parallelised “in-flight” approach
As both model creation, testing and refinement happens at the same time, all decision-making processes are interdependent by default
Links cause to effect - not through inference or correlation, but through direct connections between data input and executional outputs
PRODUCT
The way in which Disc does data creates the potential to generate exponential returns
Model Data
Quantify stochastic behaviours of each stream, and stream-of-streams, mathematically
Explore Data
Select from a catalogue of pre-curated realtime data feeds
Model Refinement
Parallelised simulation and testing environment
Build Trading Bot
Use Disc’s OpenBox AI model to build dynamic trading bots
HOW IT WORKS
How Does Disc Work?
no need to do the engineering - just do the maths
1/3
Stochastic Model
Suppose a simple Gaussian distribution to model the price of a given stock which deviates around a mean value
2/3
Real World Observations
Observations are made in real-time; we can therefore evaluate the probability of having made those observations
3/3
Trading Strategy Execution
The vectorised probability for the next observation can be formed; this is used to build an N-dimensional decision model