- Community Home
- >
- HPE Community UK, Ireland, Middle East & Africa
- >
- HPE Blog, UK, Ireland, Middle East & Africa
- >
- Financial Modelling and a Changing World
Categories
Company
Local Language
Forums
Discussions
Forums
- Data Protection and Retention
- Entry Storage Systems
- Legacy
- Midrange and Enterprise Storage
- Storage Networking
- HPE Nimble Storage
Discussions
Discussions
Discussions
Forums
Discussions
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
- BladeSystem Infrastructure and Application Solutions
- Appliance Servers
- Alpha Servers
- BackOffice Products
- Internet Products
- HPE 9000 and HPE e3000 Servers
- Networking
- Netservers
- Secure OS Software for Linux
- Server Management (Insight Manager 7)
- Windows Server 2003
- Operating System - Tru64 Unix
- ProLiant Deployment and Provisioning
- Linux-Based Community / Regional
- Microsoft System Center Integration
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Discussion Boards
Community
Resources
Forums
Blogs
- Subscribe to RSS Feed
- Mark as New
- Mark as Read
- Bookmark
- Receive email notifications
- Printer Friendly Page
- Report Inappropriate Content
Financial Modelling and a Changing World
Unless you work in the domain of quantitative finance, the term "backtesting" may well be new to you. The concept is relatively straight forward. We take an algorithm or trading strategy and apply it to historical market data. If our algorithm correctly predicts the outcome of a given scenario in our historical data, we surmise it may well do the same again when applied to real-time market data. The key phrase here is "may well", as the well-worn phrase states "past performance does not guarantee future gains". Our problem therefore is, how do we make our prediction more reliable and thus how can we confidently predict the future shape of a given financial market using a mathematical model?
The principles are the same as those used in weather forecasting. When trying to predict the weather, typically you'll find massive supercomputers like the ones from HPE Cray working round-the-clock to apply well-known forecasting models to historical weather readings taken from stations all around the world. A meteorologist will use the output from this vast computing power and analysis to publish a given forecast, often with an element of tweaking based on human experience. In finance, the process is remarkably similar.
The more historical data points you can test with, the better your chances are that your model will accurately predict the future. Thankfully in finance, we have no shortage of historical data to choose from if we can organise it in an accessible way. The theory is that when modelling financial markets, a variation on most things have happened before, and will likely happen again. Can your model react appropriately to all of these historical scenarios when chained together in a previously unseen order? Unfortunately, to brute force test using the entire back history of our financial markets data is not feasible due to the huge volume of data it would involve. Even today we struggle to analyse everything we'd like to in real time due to the sheer volume of data generated every second of every day. This presents a unique challenge for backtesting; deciding what subset of historical data am I going to test with?
But the world has changed and we no longer must test everything through brute force replay of historical market data. Increasingly we are turning to Machine Learning (ML) to build complex neural networks which we can train using historical data and then use in real time on live market data. This requires a very different compute solution, and at HPE we have been working with NVIDIA to begin to address this problem. By far the most efficient way to train a neural network is through the use of Graphics Processing Units (GPUs). We do not use GPUs for the historical use cases of rendering complex 3D images on a screen, but rather we use their highly optimised processing power to solve complex multi-dimensional mathematical problems efficiently. But even here we are limited, and a single GPU will limit the size of the model you can train efficiently, and thus we turn to systems which contain multiple GPUs all connected together at high speed so they can operate as one.
With our joint experience in Artificial Intelligence, HPE and NVIDIA can recommend new approaches to traditional mathematics that tackle what was once a brute force processing power problem. The use of a well-trained AI model as part of a back testing strategy will vastly improve the efficiency of our computations and allow us to test many more scenarios before applying a given algorithm to the live market. We have developed a range of systems that are designed purely to cope with large scale problems exactly like this. When utilising HPE Apollo systems engineered to hold the large numbers of NVIDIA Tensor Core GPUs, the scale of backtesting you can now undertake is breath-taking.
To find out more please download this brief paper and email UK_AI@hpe.com to get in touch with one of our experts.
Adrian Lovell
Hewlett Packard Enterprise
twitter.com/HPE_UKI
linkedin.com/company/hewlett-packard-enterprise
hpe.com/uk
- Back to Blog
- Newer Article
- Older Article
- Mohamad El Qasabi on: How HPE is accelerating digital transformation in ...
- MargaretN on: Welcome to the Middle East Region Community Blog
- Martin Visser on: Everything-as-a-Service: Is your organisation read...
- Kevin Barnard on: Planning for what is next – Overcoming current cha...
- Chris_Ibbitson on: Multi-cloud in Financial Services
- DJMutch on: Think global. Act circular. The circular economy a...
- BrianJenkinson on: NVMe alone is not enough, it’s time for Storage Cl...
-
Coffee Coaching
7 -
Technologies
296 -
What's Trending
62 -
What’s Trending
159 -
Working in Tech
153