Close Form

Sidebar Request Form

Sidebar Request Form

Get Free Recent Forecasts up to 86% Accurate!

Getting the Most out of Neural Networks

  • +
  • -
Neural networks start with some small baby steps, but working through numerous input/output iterations, they can recognize very sophisticated patterns to solve many real-world problems. Ted Warnock, an aerospace engineer, says you don’t have to understand the whole process built into TradeShark’s effective training program to benefit from the “amazing capability” it offers to traders.

Video Transcript

Show Full Transcript

My guest today is Dr. Ted Warnock he is a PHD in aerospace engineering, we’re talking about artificial intelligence and neural networks.  Dr. Warnock, how do you use a neural network in the financial markets to help you trade?

Well, basically a neural network can be used to bring in inputs from all the related markets that impact the market that you are interested in trading and the information from each of those markets can be used to predict what’s going to happen in the market that you’re interested in.

So can you kind of in laymen’s term define what a neural network does or what it is?

Sure, a neural network is a model of how we understand human intelligence to work.  It is composed of very simple mathematical processing elements.  Each element performs a very simple mathematical calculation however, when these are connected to each other and connected in layers, they can do some very, very powerful pattern matching and pattern recognition.

So why is it that neural networks are such a good fit for predicting financial markets?

Well, because we’re dealing with noisy and fuzzy data, basically things happen every day that makes sense in markets and things happen every day that don’t make sense in the markets.  So it’s not an exact science.  Neural networks are very good at pattern recognition in noisy data.

Alright, specifically in TradeShark, how does it give the trader an advantage in the markets?

Well, basically, first of all you have the advantage that many intermarkets are being used to attempt to predict price behavior in the market of interest as opposed to someone looking at price in the market you are trading.  That’s the first advantage, the second advantage is the power of the neural networks that have been trained to recognize patterns when certain events happen, when certain behaviors happen in the market of interest as well as the intermarkets.  The neural networks in TradeShark are able to predict what the effect will be on the market of interest price.

How powerful is it to have all this power on your desktop at home or in an office for an individual trader.

Oh, it’s an amazing capability.  You know, years ago you’d been talking super computers and huge teams of people to make something like this happen.  Today you’ve got this kind of capability in your own PC right on your desktop and all of that is transparent to the trader who’s simply interacting with a very nice price chart of the market of interest.

So speaking of teams of people, I know that TradeShark has got teams of people that have worked really hard to develop this.

Yes, absolutely, I’m sure of that, this is no small feat.

Let’s talk about specifically how a neural network works in the software.  Do I have to understand neural networks and how they work and what they do to trade well with this?

No, basically everyday market information is ingested into the software.  The neural networks process that information very very quickly, we’re talking seconds and provide their predictions.


So, neural networks and artificial intelligence are not just in the financial markets obviously but all around us in different applications.  What else have you seen?


Sure, I actually have applied neural networks to some very difficult aerospace engineering problems.  I was faced with the challenge of predicting orbital lifetimes of tethered satellites several years ago and I discovered there was no way to solve that problem without neural networks.  I was able to develop neural networks apply them to the problem and show that they worked when nothing else did.  Neural networks are used today in the financial world when financial institutions make choices about who they are going to lend money to or what your credit risk is.  There using neural networks to sort through all of your personal information, your credit history and other information about you to make a decision about lending.

One of the other aspects of neural networks is the learning or the training component.  Can you speak to that a bit?

One way I like to think about this is the way that we humans learn to walk when we’re very young.  We slowly learn to take steps and eventually we can walk on very flat surfaces pretty well if we hang on to the wall.  Later on when we’ve fully mastered the art of walking we are able to walk up hill, downhill, run, go sideways, jump do all kinds of interesting things just from that basic skill we first learned at a very elementary level.  Neural networks are very similar.  Their shown training data that consist of inputs and outputs.  Basically the network is given a set of inputs, it calculates an output.  If that output doesn’t match the desired input, the network is then given feedback so that it can adjust itself so that the output then better represents or better matches the desired output.

            So it sounds like any software really can be taught to back test well which TradeShark certainly did but it’s going forward where it really sets itself apart.

            Absolutely and what I’ve seen in my experiences is that there is an art and a science to training.  You can over train networks to the point that they will represent the training data very very well but when you try to interpolate or even worse extrapolate with the networks, they don’t do very well.  What I saw in TradeShark, when I looked at four tests data was extremely good performance across the board which tells me that the networks have been designed and trained very very well.

            There may be people out there that think that for whatever reason neural networks don’t work or can’t be applied to the financial markets.  What do you say to that?

            Well, I’d have to say that from my experience I’ve seen neural networks do some very amazing things.  Most people are familiar with the simple XY plot, right, we can plot some equation, Y equals X squared on an XY graph and we might be able to figure out the equation for that particular curve right? If all I knew were the data points.  Think about neural networks from this stand point.  What if I had 25 inputs and a single output that I’m looking for, would I even be able to visualize that as a human and figure out how to take those 25 inputs and predict or calculate that one output?  The answer is not only probably not but no and I have seen neural networks do exactly that in real world problems. Ergo the financial markets.  We’re taking about massively interconnected markets that all impact other markets.  We’re tempting to look at the way those, outside those intermarkets perform and what those intermarkets are doing and predict what our market of interest is going to do.  That’s exactly the sort of things neural networks are very very good at.

Dr. Warnock thanks for being here.

Thank you Tim.

We’ve been talking about neural networks in the TradeShark software with Dr. Ted Warnock.

Are you ready to get started?