Frederik Kryger-Baggesen
The future of quant finance is at our doorstep. The industry’s use of AI is starting to boom, and with that we have achieved great efficiencies, but what next? According to Frederik Kryger-Baggesen, Quant Analyst at SaxoBank, there’s a lot of room to improve still, and in this article, he explores the areas which would benefit the most from AI.
Smart decisions are based on great analytics, but as financial data contain a high noise-to-signal ratio, great analytics often require the processing of vast amounts of data. This limitation can be challenging when lightning quick decisions are needed.
The flourishing and on-going development of artificial intelligence (AI) supporting software and hardware solutions can in the near future show to be very beneficial in overcoming the obstacle of data hungry analytics in quant finance. The nature of deep learning models – and other types of models within machine learning – allow great accelerations in computational time by parallelising their execution onto a large number of GPUs. This great benefit has led to the development of GPU servers with efficient cores and quick GPU-to-GPU communication, build with the sole purpose of training and executing deep learning models. By acquiring financial analytics with the application of deep/machine learning models that can be executed on such servers, one can obtain a great improvement in computational power. This increase in computational power will lead to faster analytics and possibly in a way that paves the way for prior unseen possibilities within quant finance.
With this great amount of computational power available, the task from a quant perspective will be how to take advantage of it. Firstly, there is the task of obtaining useful data. The amount of observable data is enormous, but to extract and organise the useful parts is no easy job. Such cases could be the extraction of relevant information from large text documents or organising huge sets of raw financial data. To do this manually is practically not achievable, and thus appears the need of clever automated processes. Such processes can preferably be driven by machine learning models, and hence accelerated when executed on the prior mentioned GPU setups in a parallelised way.
Secondly, is the task of exploiting this extracted and organised data to make decisions on the financial market. The combination of the power of AI and the fast execution on 1000s of GPUs gives the opportunity of acquiring information-full real-time analytics that empowers the possibilities within machine automated trading. With rapid AAD sensitivities and AI real-time analytics executed on fast GPU servers, the integration of a smart and fully machine automated trading strategy to manage the risks of a financial portfolio in real-time becomes reality.
While the hardware needed to perform such analyses is already available, it should merely be matter of time before quant finance is reliant on clever automated machine learning processes that enhances both the extraction of data and the exploitation of data. This could ultimately lead to a financial market that, compared to its current state, is far more driven by automated trading executed by computers.
from QuantMinds International on
QuantMinds365