comes to picking the right kind of net to use, well, things can get a little confusing.

You first need to decide if you're trying to build a classifier or if you're trying

to find patterns in your data. Beyond that, I’ll try to help by giving you some general

guidelines.

Before we get started, I want to give you a bit of a heads up. I’m going to be using

some terminology that may sound a little scary right now, but don’t worry. I’ll cover

all these terms in detail in the upcoming videos.

If you’re interested in unsupervised learning – that is, you want to extract patterns

from a set of unlabelled data – then your best bet is to use either a Restricted Boltzmann

Machine, or an autoencoder.

What type of projects would you need to use a Deep Net for? Please comment and let me

know your thoughts.

If you have labeled data for supervised learning and you want to build a classifier, you have

several different options depending on your application.

For text processing tasks like sentiment analysis, parsing, and named entity recognition – use

a Recurrent Net or a Recursive Neural Tensor Network, which we’ll refer to as an RNTN.

For any language model that operates on the character level, use a Recurrent Net.

For image recognition, use a Deep Belief Network or a Convolutional Net.

For object recognition, use a Convolutional Net or an RNTN.

Finally, for speech recognition, use a Recurrent Net.

In general, Deep Belief Networks and Multilayer Perceptrons with rectified linear units – also

known as RELU – are both good choices for classification. For time series analysis,

it’s best to use a Recurrent Net.

Deep Nets are the current state of the art in pattern recognition, but it’s worth noting

that neural nets have been around for decades. So you might be wondering: why did it take

almost 50 years for Deep Nets to come on to the scene? Well, as it turns out, Deep Nets

are very hard to train, which we will see in the next video.