Quantum Deep Learning
arxiv.org
arxiv.org
The first approach considered in the algorithm really doesn't need a quantum oracle. Since that algorithm only requires that you sequentially feed the input data you want to train the system with, you don't need any sophisticated quantum algorithm to provide an entangled mixture of the data used to train the system.
The second algorithm we consider does use a quantum oracle (meaning a quantum device that allows the quantum computer to prepare an entangled mixture of input training examples). In practice, if you have a database of training vectors (like MNIST handwriting images that you want to recognize) then you'd need to make a giant quantum computer that stores all of this and allows quantum access to this. This could be done using a QRAM (although not all quantum computer designs have efficient QRAM).
Alternatively, the quantum oracle could be any other quantum algorithm that you want to learn about. One idea that I'm a little obsessed with is the idea of using a quantum simulation subroutine as the quantum oracle. Then this approach allows you to train a deep Boltzmann machine to learn features of the system that the subroutine is simulating. Ideas like this could really accelerate drug testing and development by using AI and quantum simulation simultaneously to focus in on promising candidate drugs without requiring as much trial and error as current methods.
I actually did some preliminary investigations, but unfortunately using factor graphs as a way of quantum state is apparently an area of active research. That makes searching for using quantum computers to calculate factor graph traversal pretty hard!
This paper looks pretty interesting though.
"Research" is probably an exaggeration, it was mostly some Googling on a Friday afternoon, and some whiteboard scribbles and reading the D-Wave docs.
Some of my colleagues have talked to D-Wave previously, so we are only mostly ignorant.
The extra, "Fuzzy," makes it better. With two data points, you get sets of attributes, ad infinitum. Rather than trying to simply point out what ones are common, or good, or match any given specific outcome, etc. I like infinitely awesome attributes.
Tastes like bacon.
Have you ever tasted the difference between fuzzy logic influenced design and the probabilistic stuff? Infinitely more delicious.
Case in point: sometimes, you simply want to know an expectation was met - or not; and that there was the initial "expectation." Once this becomes clear, the chain of trust is easily, "fuzzed," to get some interesting outcomes.
Research a "three dimensional model of taste," which doesn't exist (yet). Why? Think of a six sided dice, every facet, which combined with it's twin, adds up to seven, a prime.
Now, extrapolate from the flavors you've enjoyed the following: 1. Sweetness - strawberry perhaps? 2. Sourness - a lemon? 3. Saltiness - a saltine cracker maybe? Thirsty? 4. Bitterness - funny how the lips pucker from Grapefruit. 5. Umami / savory - we might not officially recognize this in some places, but we all know "savory," - potatoe soup's a favorite of mine. 6. Spicy - are you feeling the heat, yet?
Now, compare that to any given model of flavor...I'm waiting. Thanks :)
It doesn't match, does it? With taste & smell, alone, a human being can sense four dimensions...all three axes, X, Y and Z on the Cartesian plan. Oddly enough, most flavor models left out, "Spicy," and after I saw a German guy's head nearly explode from a Thai Chili Pepper, I knew, *spicy exists
What's the fourth, for the person who can't see, who can't hear, who can't talk?
Time. If you can taste, taste changes over time. If you can feel, you're Hellen Keller and can understand Braille.
Now, Helen Keller - please next time, throw me some of your clearly larger, more infinite, more awesomely epic knowledge, because I'm confused.
I'm also blind, can't hear, and can't see. When you sh!t on my, "Karma," and refuse to share your overwhelmingly superior knowledge...
...well, we all know :)