An autoencoder is an artificial neural network where output units are directly connected back to input units, and is often used for unsupervised learning of efficient codings.

In finance, autoencoders and other deep learning methods are applied to large data sets with complex interactions to produce more useful results than the industry's more traditional methods.

Some experts believe financial markets are inherently inefficient, and are becoming more and more inefficient with time. More complicated models may be needed to capture dwindling opportunities in the future, and deep learning could be the key to helping companies make better predictions for success in the markets. Shivaram Ramegowda, formerly AVP at Societe Generale, leading the cross asset quantitative team in applying machine learning and time series analysis, allowing them to build better models for financial market forecasting. At the Deep Learning in Finance Summit on 27-28 April, Shivaram will share insights on the applications of deep learning algorithms in finance, such as autoencoders, and explore their value over conventional methods.  I asked him a few questions ahead of this month's summit to learn more about his work, and what we can expect next for deep learning in the financial world. How did you begin your work in deep learning?I have worked extensively on SVMs and when I was doing my Masters about 12 years ago, people had written off neural networks. SVMs were the cutting edge. Recently, I started noticing shape improvement in performance of machine learning algorithms on images due to deep learning. When I also saw similar performance improvements on other data, I wanted to give deep learning a try. Deep learning also made sense as I believe in the power of evolution and deep learning is inspired by it.  Which industries or areas do you feel deep learning will have the most beneficial impact?Deep learning is highly scalable. Any area which has a huge amount of data will be benefited by deep learning (almost all the areas will be impacted).  What advancements in deep learning would hope to see in the next 3 years?I hope we can build models which can understand the context in text and time series. We have now built models which work very well on one kind of data, for example, image classification, OCR, speech to text etc. But the financial data consists of time series, text, images, and other formats. I hope we can train a single model using all kinds of data (Generic model). Can we simulate the entire market in a giant model and train an algorithm using reinforcement learning?  Do you think deep learning will continue to influence financial technology, or will it have more impact on other sectors?Deep learning already has huge impact on financial and other sectors. I don't think that will change.   What do you think is the next advancement beyond deep learning?Deep SVMs or Deep Trees? We may use different basic units to build the network. Better models which can learn with small amount of data. Generic models which can learn any problem.There's just 2 weeks to go til the Deep Learning in Finance Summit in Singapore! The summit will take place alongside the Deep Learning Summit on 27-28 April, view further information here.

Other confirmed speakers include Sonam Srivastava, Quant Analyst, HSBC; Ilija Ilievski, PhD Student, National University of Singapore; Edouard D'archimbaud, Head of Data & AI Lab, BNP Paribas; Siddhant Tiwari, Data Scientist, AXA Data Innovation Lab; and Scott Treloar, Founder, Noviscient.

Tickets are limited for this event. Register your place now.

The Deep Learning in Finance Summit will also be held in London on 1-2 June and Montreal on 12-13 October. View all upcoming events here.Opinions expressed in this interview may not represent the views of RE•WORK. As a result some opinions may even go against the views of RE•WORK but are posted in order to encourage debate and well-rounded knowledge sharing, and to allow alternate views to be presented to our community.