Don't read if you're not interested in deep learning
Why are there neural networks and deep neural networks?
Hey friends, hoping your day got off to a good start. In this zap, I want to take a quick dive into the world of data to discuss deep learning.
Soo..what exactly is deep learning?
Let's all be data scientists for a second. When we think about training machine learning models in general, there's a lot of work that needs to be done before we can apply any kind of algorithm.
To summarize, input data first needs to be cleaned up and features need to be created. A feature is a characteristic of the data that can be used for further analysis; if we looked at the passengers on the Titanic for example, some features would be their age, fare, or even the deck their cabin is on. Data typically comes with features like these right away, but oftentimes you will want to create your own so that your model can be more accurate.
Okay, we’ve got that out of the way, onto the neural networks!
Neural networks are special because they combine the steps of feature extraction and classification for you. A neural network will have an input layer that you feed your data into, a hidden layer where it tries to learn the most important features of the data on its own, and an output layer that provides the prediction (in our Titanic case, whether the passenger survived or not).
A deep neural network (DNN) would be a model that contains multiple hidden layers. At this point, the structure begins to resemble that of an actual brain — each layer can learn something new about the original data to ultimately make a prediction about it.
These can get really complicated and some architectures (which is a fancy way of describing how many nodes and layers a neural network has) like ResNet can be hundreds of layers deep!
So the next time you're wondering how Amazon knew exactly what product to recommend for you, there's probably a DNN out there that used data from thousands of shoppers like you to make a prediction about what you are most likely to buy next.
In this deep dive Thursday, we shed a little light on what mysterious neural networks actually do. Do you have an idea for an app that uses a neural network in some way? Perhaps you're thinking of making an MVP and don't know where to start. Well, check out our Prototyping with Figma cohort on Enlight if you haven't already. No experience is necessary if you want to learn how to make hi-fi prototypes that pop!
That's it for this zap, see you later!
Maxim