Page 192 - AI Ver 3.0 class 10_Flipbook
P. 192
Supervised Vs. Unsupervised Learning
The differences between Supervised and Unsupervised Learning are as follows:
Aspect Supervised Learning Unsupervised Learning
Uses labelled datasets with input-output Uses unlabelled datasets without
Data
pairs (e.g., images labelled as "cat" or "dog"). predefined categories.
To predict the output for new, unseen data To explore data and find hidden
Goal
based on the labels provided during training. patterns or groupings.
Common Techniques Classification, Regression Clustering, Dimensionality Reduction
- Predicting house prices based on features - Grouping customers into segments
Examples like area and location. based on their purchase behaviour.
-Classifying emails as "spam" or "not spam". -Detecting anomalies in network traffic.
Predicts specific labels or values for new Groups data into clusters or finds
Output
data. patterns, without specific labels.
Reinforcement Learning
Reinforcement Learning is a type of machine learning where a model learns through trial and error to make the
best decisions in a given environment. It interacts with its surroundings, receiving rewards for correct actions and
penalties for mistakes, gradually improving over time. The goal is to maximise cumulative rewards by learning from
experience and adapting to new situations.
For example, a robot learning to walk starts with random movements. When it takes a correct step, it gets a reward;
if it falls, it receives a penalty. Over time, the model learns which movements keep it balanced and eventually walks
efficiently.
Let us take an example of Reinforcement Learning:
You show the machine an image of a ball and ask it to predict the object. Initially, the machine predicts it to be a
globe. Since this is incorrect, it receives negative feedback. The machine then adjusts its understanding and learns
that the object is not a globe. When the same image is provided again, the machine predicts it to be an orange.
This is also incorrect, so it receives negative feedback again.
Reinforcement Model Globe Reinforcement Model Orange
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
190 Touchpad Artificial Intelligence (Ver. 3.0)-X

