Will QLattice ever rule the World?

qlattice machine learning
To answer that question, yes, or atleast I hope so. Because that would mean making life so much easier for Machine Learning Engineers, enthusiasts, and beginners equally. How? Well, I'll tell you that in a bit. But first, let me tell you what is QLattice.

QLattice is the Machine Learning environment that takes into account all the possible models of a dataset and finds that 'path of least resistance', or in other words it finds the model that best fits the problem. The models looks somewhat similar to a standard neural network, but there are differences. Now let me tell you how it works.

So how does it work?
Let's say you have a dataset that you would like to model using Feyn and a QLattice. Feyn is the SDK that is used to interact with the QLattice, it is named after Richard Feynman because of his work on path integral formulation which was an inspiration for QLattice. Feyn is pronouced like "fine". First, she will extract a QGraph from a QLattice.

What on Earth is a QGraph?
QGraph is an unbounded list of potential models for the dataset. Here's a graph from the Airbnb dataset regression model:
qlattice airbnb regression
Looks somewhat like a Tensorflow graph, doesn't it?

Understood. Then what?
Next she will fit the data on a subset of these models and update the QLattice with the best one. This basically tells the QLattice that this was a good model, can we have more of this model please! Now, the search space of the QLattice narrows down and focuses in this direction. The next time she fits the QGraph it will be on more relevant models for the dataset.

Isn't that great! But wait there's more. With QLattice, there is no need for you to one-hot encode any of the categorical features. If a feature takes only categorical values then feyn automatically encodes them. You don't need to scale any of the numerical features because it's automatically scaled to be between [-1,1]. And, to top it all, QLattice makes model interpretability the new normal. If that doesn't makes your life easier, I don't know what will.

Want to know more?
Visit their website. Want to do some tinkering with it? Well go right ahead, they have a bunch of Jupyter notebook and quick start guides available for your to play with. However, to do that you need to have your own QLattice URL, and a QLattice Token. Afraid not, there are some cool people out there at Abzu (the creators of QLattice). Just ask them politely, and I am sure they will help you get an early access.

Here are the basic requirements:

        Debian-based Linux, Ubuntu 18.04+
        Mac OS X 10.14+
Python 3.6, 3.7 or 3.8

Well, what are you waiting for? Take it out for a spin today, and have fun!