About Us
The firm specialises in finding causal relations in heterogeneous data and making verifiable predictions based on them. For us, the unique power of Machine Learning lies in that it may help us see more clearly as we take a higher perspective at larger and larger volume of data.
What is on Our Mind
How to Forecast
We consider that any two objects could boast of predictive power for each other, a pull force by "information gravity". The problem is to compute the conditional distribution of B's change in the future given A's change at present.
- Koenker, R. (2005). Quantile Regression (Econometric Society Monographs). Cambridge: Cambridge University Press. doi:10.1017/CBO9780511754098
- Jürgen Schmidhuber, Artificial Curiosity & Creativity Since 1990-91.
How to Reason (or Imagine)
How to program programming? In this realm, most problems are computationally hard. Furthermore, one may start to construct ideas that only exist in the mind, from known premises.
- Sterling, Leon S. and Ehud Y. Shapiro (1994). The Art of Prolog, 2nd Ed. MIT Press.
- Colas, Cédric and Karch, Tristan and Moulin-Frier, Clément and Oudeyer, Pierre-Yves (2021). Language as a Cognitive Tool: Dall-E, Humans and Vygotskian RL Agents. tweet.
Highly-Vectorised Data Processing
We would generally favour a simple, transparent, and efficient hardware/software architecture for processing data.
- RISC-V: an open ISA.
- JAX, a Machine Learning framework.
- McSherry, Frank (2015). Differential dataflow.
Contact
Email: info AT brief-ds.com