Hi,
A human brain is 1400cc in size and 100 000 times slower than a modern computer. Yet Gary Kasparov managed to beat Deep Blue quite a few times since their first encounter in 1996. Even though the Word Champion has been defeated more recently, a human being has still inflicted severe defeats to the most advanced computer of its time. Impressive when we all know that computers are smarter, faster and better looking than we are? 
Why is that?
Let’s say that modern computers are nothing more than giant calculators without a brain: they just calculate series of equations without too much of interaction within the equations. Such an architecture requires CPUs to be fast. They also consume a lot of energy and generate a lot of heat (sometimes leading you to unexpected reboots)
A human brain still is 10 billion times more efficient than a computer in terms of calculations and power consumption. Indeed, we rarely generate enough heat to make our brain to reboot. We also do not experience a kernell32.dll or a “blue screen” fatal error either (mostly because we don’t use Windows… )
How does it work?
Since I am not a scientist I’ll make it short: a human brain uses 6 or 7 layers of neurons for a total of 10^11 neurons that work as a “parallel computer” that learns from the feedback of other neurons and changes its structure or creating new connections with neurons when required. In brief: the brain adapts to a changing world…
How about combining the best of both worlds in Excel then?Humans can forget, a computer doesn’t.
A computer can replicate 100 million times the same task with a limited risk of errors.
So now, a bunch of companies have designed softwares to teach our computers to act as a human being (scary…)
Benefits of Neural Networks in Excel:
From information to knowledge their is sometimes as little as classification.
Just sorting your data in the most efficient way gives you an edge. Ok: Excel comes with a tool to sort data. But using Neural Networks can be faster and more more efficient as shown here: data classification with classifierxl
You can sort large excel databases, create clusters and generate charts in a breeze.
You can also forecast data series or perform backtestings.
Standalone products also exist.
You can also build “better”(their claim, not mine) investment strategies and portfolio allocations. Look here.
For professionals in the Credit Industry, credit checks using decision trees and neural networks can be a good opportunity to better assess the creditworthiness of their clients using this tool.
How does it works?
In the above graph you can see the first layer of neurons on the left (the inputs), then an additional layer and then the output. In the case of an investment strategy, let’s say you want to forecast the behavior of a stock.
the inputs would be: open/close/high/low/volume
the 2nd layer would be the technical analysis indicators you usually use (RSI, Bollinger’s bands etcetc)
the output would be the closing price.
In my opinion Neural Networks is the new trendy thing that will be useful in the future for very complex environments and AI based computers. I think that a neural networks add-in in Excel will do the same job as another add-in in the same league using another classification algorithm. It will just perform the task in a different way.
Yet it still is interesting to know that some add-ins a very easy to use to classify a large database in Excel and maybe help in building trading strategy!
So: what’s your view? Are neural networks the next big thing or are they just a fad?
