Page 126 - Ai_C10_Flipbook
P. 126
Finally, after multiple attempts and adjustments, the machine
predicts the object correctly as a ball. Positive feedback is given,
and the machine has successfully learned to identify the ball. Reinforcement Model Ball
Other examples of Reinforcement Learning are self-driving
cars, robotics and a variety of video games available these days.
1. Video Game AI: A video game character (AI) learns to
play a game like chess, Go, or Atari games. In video game
AI, Reinforcement Learning allows agents (e.g., characters or bots) to learn optimal strategies by exploring
the game environment, receiving rewards for achieving objectives, and penalties for failing, improving their
performance over time.
2. Car Parking: In car parking, Reinforcement Learning helps the car learn how to park properly by trying different
ways, getting rewards when it parks correctly, and getting penalties if it makes a mistake like hitting something.
3. Humanoid Walking: Humanoid walking is related to Reinforcement Learning as the robot learns to walk by
receiving rewards for stable movements and penalties for falls, adjusting its actions to maximise balance and
efficiency over time.
What Makes it Different?
Reinforcement Learning is unique because of its ability to handle situations where traditional methods like
supervised or unsupervised learning may fall short. In these approaches, you typically need a clear understanding
of the data and the problem you're solving. However, real-world scenarios often involve complex and dynamic
environments that are not fully understood, and data might be lacking in certain situations. Additionally, the
environment may change over time, requiring the system to adapt. Reinforcement Learning stands out because it
doesn't rely on extensive pre-existing knowledge or large datasets, allowing it to learn from interactions with the
environment and adapt to unforeseen circumstances, making it highly effective in dynamic & uncertain settings.
Difference between Supervised Learning, Unsupervised Learning and
Reinforcement Learning
Aspect Supervised Learning Unsupervised Learning Reinforcement Learning
Nature of Labelled data with Unlabelled data; no No labelled data; the model learns through
Data input-output pairs. predefined output labels. interaction with the environment and
receives rewards or penalties for actions.
Learning Learns a direct Learns patterns, clusters, or Learns to take actions that maximise
Process mapping between input structure in data without long-term rewards based on trial and error.
and output. predefined labels.
Feedback Instant feedback based No explicit feedback, learns Feedback is delayed and occurs based on
on the correct label patterns or relationships. the outcomes of actions.
Sub-categories of Supervised Learning Models
The Supervised Learning is further categorised as: Classification and Regression. Let us learn about these in detail.
Classification
Classification is a rule-based AI model. It is a systematic grouping of observations in classes, something like
categorising plants, animals in different categories by biologists. In classification you teach the machine to perform
with labelled data. Testing data is then classified as one of the labels of the training dataset.
124 Artificial Intelligence Play (Ver 1.0)-X

