Like most industries, artificial intelligence is expected to impact the financial sector in numerous areas, including blockchain technologies, robo-advisors and virtual banking assistants, financial forecasting, stock market prediction, and fraud detection among others.
Deep learning in particular has been making great advancements in finance over recent years. Financial prediction problems, such as pricing securities and risk management, usually involve large data sets with complex interactions, making it difficult or impossible to make use of the data in a full economic model. By applying deep learning methods we can produce more useful results than the finance industry’s more traditional methods.
Scott Treloar, Founder of Noviscient, will be joining us at the first Deep Learning in Finance Summit in Singapore next month, to share expertise on the use of autoencoding to learn about latent factors that drive security returns. I spoke to him ahead of the summit to learn more.
Can you tell us a bit about yourself and your work? Plus, give us a teaser of your presentation at the summit?
I have a varied background that includes venture capital, trading, risk management and software development. I recently founded Noviscient here in Singapore. We consider ourselves a next generation investment manager. We are a technology company operating in financial services. We source systematic trading strategies globally and execute the trading locally. We operate a modern, cloud-based systematic technology platform that has been developed here in Singapore. We make full use of machine learning and open source software. We work with state-of-the-art messaging, database and visualisation technologies.
How did you get into deep learning?
While I initially programmed in C++, R and Matlab, I switched over fully to Python around eight years ago. I note that a very large systematic fund manager similarly a collection of languages including R, Matlab and Java to do everything in Python. So I think I am in good company.
In systematic trading, you are trying to predict future returns and/or you are trying to construct portfolios of securities with certain characteristics. As I was doing my work in Python, I naturally started to use some of their machine learning libraries such as scikit-learn for prediction. Over time we began to realise that the linear relationships between predictor and response variables had been mined away due to the competitive nature of trading — everyone was doing regressions. This led to our work with non-linear relationships for prediction. Deep learning is a good framework for identifying and exploiting non-linear relationships between predictor and response variables.
Which industries or areas do you feel deep learning will have the most beneficial impact?
Too many to talk about. The confluence of big data, cloud computing and advancements in deepl learning algorithms represents a massive change in how the world will be operating. While we all tend to think linearly, the combination of these factors will have an exponential impact on the world.
What advancements in deep learning would hope to see in the next 3 years?
We are keen to see more theory-based work done with deep learning on ordered data sets, specifically time series. Often time series are time-inhomogeneous. That is, the underlying generative process for univariate time series is changing. And also the dependence within multivariate time series will be changing. In these situations, the usual static data assumptions of more data being monotonically better don’t necessarily hold.
Are fintech startups likely to be a large factor in the adoption of Deep Learning?
FinTechs are not burdened by legacy infrastructues and organisational practices — which greatly inhibit incumbents. FinTechs, such as Noviscient, are born cloud-native, with the ability to use the best of open source technologies and partnering approaches to create value for their customers. I would expect FinTechs to be at the front of innovative use of deep learning technologies.
Scott Treloar will be speaking at the Deep Learning in Finance Summit, taking place alongside the Deep Learning Summit in Singapore on 27–28 April. Learn about deep learning applications in the financial sector from algorithms to forecast financial data, to tools used for data mining & pattern recognition in financial time series, to scaling predictive models, to stock market prediction, to using blockchain technology. View further information here.
Tickets are limited for this event. Register your place now.