Productionizing AI by 2023

Author:2023
Language: eng
Format: epub

Chapter 5 Neural Networks aNd deep learNiNg

Figure 5-10. LSTM Architecture

Other Types of Neural Networks

Convolutional and Recurrent Neural Networks are used as deep learning architectures to

train on Supervised Learning problems, that is, where we have prelabeled data (such as

classified images or end-of-day stock prices for forecasting). This is the focus of our first

hands-on exercise at the end of this section.

Before we go to that, we will take a quick look at other architectures which are

generally applied to solving Unsupervised Deep Learning problems. These include

Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), and Deep

Boltzmann Machines (DBMs) as well as Autoencoders, Variational Autoencoders,

and Generative Adversarial Networks. Autoencoders are addressed in our second

hands-on lab in this section.

Restricted Boltzmann Machines (RBMs)

RBMs are generative, stochastic two-layered artificial neural networks which learn a

probability distribution over a set of inputs. There are only two types of neurons, hidden

(h in diagram) and visible (v in diagram), all of which are connected to each other. There

are no output nodes.

Whereas Boltzmann Machines have connections between input nodes, Restricted

Boltzmann Machines are a special class with restricted connections between the visible

and the hidden units. This allows for more efficient training using gradient-based

contrastive divergence algorithms.

149

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.