Automatic trading, although it has reached a certain technological maturity, still demands significant investment in R&D. The work may traditionally focus on stochastic mathematics, but certain quantitative American investment funds – e.g. Renaissance Technologies, Two Sigma, Winton Capital and Man AHL - have been interested for several years in machine learning.
Following their example, Kevin Lourd and his seven partners at Walnut Algorithms have developed an approach based on deep learning and reinforcement learning. Deep learning is a form of artificial intelligence that uses the multi-layer neural networks, and allows for the creation of algorithms with several levels of abstraction. It’s the basis for the latest image recognition tools and automatic translation software.
Learning through reinforcement is founded on algorithms capable of learning automatically, but also of evolving by optimizing a reward function based on past experience. One such example recently succeeded in beating Europe’s current go champion. Like humans, the algorithms can adapt to external conditions, learn from their mistakes and anticipate a situation. Unlike them, they have almost limitless memory and computing capacity – characteristics that make for excellent agents on the financial markets.
This is what Walnut Algorithms is betting on, and they are clear about their ambition. Incubated at ESSEC Ventures, accelerated by Startup Bootcamp Fintech and recognized by Finance Innovation, they are included in the International Business Times “Watch list” 2016. The startup’s objective is to seek support from an investment fund at the start of 2017 and open an internationally renowned research center, employing around thirty engineers within five years.
Find out more: