Machine-Learning can be incredibly useful in Finance. BG approach to it is based on two pillars: Resilience and Memory.


Resilience We look at within peer-group relative performance resilience: each firm is scored on a case-by-case basis depending on its resilience. BG approach is to focus on identifying stable out/ underperformers from market behaviours in its scoring rather than relying on fundamentals or financial ratios.
Memory We focus on within peer-group representative investor memory: acknowledging the fact that investors adapt their reference point and risk aversion to evolving market conditions.

Our approach is illustrated in a Position Paper : “What can Machine Learning bring to investing?” Dr. Arnaud de Servigny & Elyes Ben Bouzid, March 2016

The current set of techniques used in the Research Lab.

Core techniques used Usage Reference
Variational Inference Looking at choices under uncertainty as approximated and simplified probability ranges
Deep Learning Building several layers of filtering to strengthen signals contained in the data
Customized learning strategies
  • Reinforcement Learning to focus on areas of higher certainties
  • Adversarial Learning to test the quality of models