Page 181 - AI Computer 10

Actions (t)

Environment

AGENT

Reward (t) Reward (t + )

Observation (t) Observation (t + 1 )

3. Rewards often contain only partial information. A reward like a win in chess conveys that some inputs must
have been good, but it doesn’t clearly signal which inputs were good and which were not.
4. The system is learning an action policy for taking actions to maximise its receipt of cumulative rewards.
Thus, Reinforcement Learning adopts an iterative approach where the AI model performs some actions, receives
feedback on its actions, evaluates the feedback, and learns from each subsequent iteration.

Difference between Supervised, Unsupervised, and Reinforcement Learning

The main difference between the three machine learning models are:
Feature Supervised Learning Unsupervised Learning Reinforcement Learning

Data Type Labelled Data Unlabelled data Interactive Data
Objective Predict output on the basis Identify patterns or Maximise rewards
of input features groupings

Learning Process Uses a training set with Learns from the data Learn through trial and
input-output pairs itself error

Applications Image classification, Email Customer segmentation, Game Playing, Robotics
filtering Anomaly detection Navigation

TYPES OF SUPERVISED LEARNING MODELS

Supervised Learning models can be broadly categorised into several types based on their purpose and the nature
of the output. The main types of supervised learning models are:
u Classification model u Regression model

Supervised Learning Models

Classification model Regression model

47
47