Deep Learning with Tensorflow - Introduction to Unsupervised Learning Video Lecture | Tensorflow: Learning made Easy - AI & ML

Hello, and welcome!
In this video, we will provide an overview
of the basic concepts of unsupervised learning.
Depending on the goals of the application,
there are many different techniques for training
a neural network.
A few general categories are supervised learning,
reinforcement learning and unsupervised learning,
which is the topic of this video.
But before explaining what unsupervised learning
is, it might be helpful to recall how supervised
learning works.
In a supervised learning algorithm, the machine
receives a dataset containing inputs and the
targets associated with each input sample.
After processing the data, the resulting output
is compared with the target from the dataset.
If the results differ, the net is fine-tuned
until it can more accurately model the training
data.
But it might be the case that we don't have
any output targets at all.
This is when we would apply unsupervised learning
techniques.
With unsupervised learning, the network only
receives the input data.
There are no output targets, or rewards, or
punishments for its results.
The goal of unsupervised learning is to find
patterns in the data, and to build new and
useful representations of it.
Unsupervised learning is a powerful technique
for tasks like pattern recognition, data clustering,
object recognition, feature extraction and
data dimensionality reduction, just to name
a few examples.
Depending on the nature of the data and the
goals of the application, there are several
different techniques to choose from.
Some of these are Restricted Boltzmann Machines,
autoencoders, Self-Organizing Maps, Principal
Component Analysis and k-means.
By now, you should understand the main ideas
behind unsupervised learning, and how it differs
from supervised learning.
Thank you for watching this video.