2019-12-16
It's Monday today, and Monday means many things to different people. I'm kidding, actually, because we all know Monday means "suck" in about 30 different languages last I checked.
But the silver lining? Rina Dechter got up and went to work on a Monday in 1986. Because of that, she added "Deep Learning" forevermore to the lexicon of science involved in machine learning.
What is it then, you're asking? What's so deep about this learning? If left to its own devices, would its poetry be so heartfelt as to surpass Shakespeare?
No, not exactly. What we call the "deep" in deep learning is really just a description of the topography. That's it. Let's get into it. Just to give you some rope to hang yourself with at your next Christmas party.
Mainly, artificial neural networks a la deep learning rely on multiple "layers" of interpretating input data. Predictive juice flows from left to right, layer by layer, slowly getting closer to a given assessment.
These layers are calibrated by exposure to training data. Meaning they actually learn themselves to represent useful slices of prediction on their way to the final layer.
This assessment is even correct sometimes, but we'll save that for another day.
What does all this mean for you, my humble reader of algorithmic non-fiction?
Well, I'm not completely convinced this particular branch of machine learning is all that useful for quant trading, at least not applied traditionally. I'll get into why in future lessons.
But fear not, getting to grips with this monster in the best way possible is still going to help you tremendously, so check out Andrew Trask's tome of awesome below:
https://www.indicativequant.com/offers/grokking-deep-learning