Using deep learning to trade

practical lessons for technologists

qplum.co

Why use Deep Learning?

The problem

Assets managed globally: $164 trillion

 

Fees charged : more than $1 trillion a year

 

Yet, we are no closer to solving the problem

qplum.co

Towards a science of investing

In future, we should be investing with a trustworthy tool and not experts.

 

The tool should look at every aspect of the data.

The tool should be affordable and efficient.

The tool should know what we have learned already."

- Benjamin Graham

... and it should keep learning

The answer is obvious ... Deep Learning

How do we use DL at qplum?

Kinds of DL that we found useful and blind alleys that we have gone down

A better relative value trade

A lot more unsupervised learning

qplum research report of who is using DL in trading

A better global macro trade

Understanding the yield curve

.. the way humans do

Technical aspects of DL at qplum

qplum.co

qplum.co

Technical and operational aspects

Supervised / unsupervised

Number of layers

What sort of data

Key bottlenecks in training and inference

qplum.co

Software hardware and deplopyment

  • Keras / Tensorflow / Caffe 2

  • GPUs / CPUs - how many?

  • Instance type and Devops

Business drivers for DL in Trading:
Why now?

qplum.co

1987, 1997, 2007, 2017?

qplum.co

Why now? - Lots of data

  • High Frequency trading and systematic trading, in general, has led to a lot of data. We are generating more data in one day now than we were in the entire decade of 1990s.

  • The traditional quant approach does not spend as much time in discarding noise.

qplum.co

Why now? - Hardware and Software optimized for DL

  • GPUs and customized hardware that allows us to solve problems in hours that would have taken weeks a year or two ago.

  • Software like Tensorflow/PyTorch and MapReduce make all of this cheap enough for small companies to innovate with.

qplum.co

Why now? - ML in social sciences

  • Trading is a social science and until recently all machine learning was focused on pure sciences.
    Deep Learning is perfect for social sciences.

qplum.co

Why now? - ML is better than traders

qplum.co

Why now? - Because of us

  • Availability of talented engineers in DevOps, Data Infrastructure and Machine Learning who can make it happen, who want to make inroads into this last bastion of inequality, who want to stop people from selling crap to investors.

qplum.co

How is it different from the traditional quant approach?

The next ten years - Deep Learning

Why is Deep Learning better than ML?

qplum.co

The traditional quant approach

  1. Hire lots of quants.

  2. They all think of trading strategies.

  3. They backtest them

  4. The firm invests in the strategies that have the best returns.

qplum.co

Problem: Too much data at every step

  • This requires hiring a lot of quants

  • They will then make millions and billions of features.

  • Challenge then is to pick the needle in a haystack of trading strategies, with very little data.

qplum.co

Investing is a science, a utility, not a game or competition, but a genuinely inclusive process

Email: contact@qplum.co  Ph: 1-888-QPLUM 4U

qplum.co

qplum LLC is a registered investment adviser. Information presented is for educational purposes only and does not intend to make an offer or solicitation for the sale or purchase of any specific securities, investments, or investment strategies.  Investments involve risk and, unless otherwise stated, are not guaranteed.  Be sure to first consult with a qualified financial adviser and/or tax professional before implementing any strategy discussed herein. Past performance is not indicative of future performance.

Using deep learning to trade: Practical lessons for technologists

By Gaurav Chakravorty

Made with Slides.com

Using deep learning to trade: Practical lessons for technologists

What kinds of deep learning have we found useful at qplum? Technical and operational aspects of trading using deep learning. Key bottlenecks in training and inference. Software frameworks and which hardware platforms have proven most useful for those workloads. What does a deployment look like? What are the scaling challenges and key drivers of cost? How does devops work when a lot of the dev is handled by machines.

  • 1,706

More from Gaurav Chakravorty