Page 107 - Code & Click - 7

For example, if the goal of the AI system is to correctly classify the objects in different images as
either ‘cat’ or ‘dog’, the labeled training data would have image examples paired with the correct
classification label. Supervised Learning systems can also be used for identifying the correct labels of
continuous numerical outputs.

Unsupervised Learning
Unsupervised algorithms are those that can
extract features from the data without the need
for a label for the results. The AI model produced
by an unsupervised algorithm would not return
that a specific input image was of a ‘cat’ or a ‘dog’.
Rather, the model would sort the training dataset
into various clusters based on their similarity. One
sorted cluster might be the desired groups of cats
and dogs, but images might instead be clustered
based on undesired categories such as whether or
not they have a blue sky in the background, or a
wooden floor.
Unsupervised Learning systems are therefore
often less predictable, but because unlabeled data
is almost always more available than labeled data,
they remain critical.

Semi-Supervised Learning
There are ‘Semi-Supervised’ algorithms that
combine techniques from Supervised and
Unsupervised algorithms for applications with
a small set of labeled data and a large set of
unlabeled data.

Reinforcement Learning
In Reinforcement Learning, the training
data is collected by an autonomous,
self-directed AI agent in the course of
perceiving its environment and performing
goal-directed actions.

Four aspects of Reinforcement Learning,
notably distinct from Supervised and
Unsupervised Learning, are:

1. Data is gathered by the AI agent itself in the course of its interaction with the environment
and perceiving stated changes. For example, an AI agent playing a digital game of chess makes
moves and perceives changes in the board based on its moves.

105