AI

Question 1

How can AI be used to Solve Machine Learning Problem?

Answer 1

• Automatically create models from data to perform certain tasks through machine learning
• Not Guaranteed perfect model, but will find good model depending on difficulty of problem
• Good for problem where it is difficult to create good models manually
• Good for problems that don’t require perfect answers

Question 2

AI in Optimisation Problems:

Answer 2

• Solve them in a reasonable amount of time through optimisation techniques
• No guarantee to find optimal solution in reasonable amount of time but a good solution
• Good for problems where no specific technique exists that guarantees that optimal solution can be found

Question 3

Think Humanly?

Answer 3

-Can machine think humanly?
- Can we consider machine as human?
- How to define think humanly?
==> With AI we need to define things with mathematical forms

Question 4

Act Humanly?

Answer 4

• What can humans do?
• What if human’s action is wrong?
  ==> Doesn’t mean we shouldn’t copy the wrong actions

Question 5

Think Rationally?

Answer 5

• Think Logically?
• Logical AI?
• Too Narrow?

Question 6

Act Rationally?

Answer 6

==> Rationality: Doing the right thing can be mathematically defined & General enough, linked to human behaviour

Question 7

Define AI using ‘Rational Agents’:

Answer 7

Rational Agents are computer programs that perceive their environments & take actions that maximize their chances of achieving best EXPECTED outcome

Question 8

What is Machine Learning?

Answer 8

An agent is learning if it improves its performance after making observations about the world.

Machine Learning when the agent is a Computer

Question 9

What is Machine Learning Problem?

Answer 9

Problems that require a model to be built automatically from data e.g => to make classification

Question 10

What are the forms of Machine Learning?

Answer 10

• Supervised Learning
• Unsupervised Learning
• Reinforcement Learning

Question 11

What is Supervised Learning?

Answer 11

• Most popular form in Real World
• Learning with a Teacher
• Teacher: expected output, labels, classes, etc.
• Solve 2 types of Problems : Classification & Regression Problem

Question 12

What is Classification Problem?
{Give example}

Answer 12

• Predict categorical class Labels

==> Spam Detection

Question 13

What is Regression Problem?
{Give example}

Answer 13

• Prediction of a Real Value

==> Student Grades, Stock Price Prediction

Question 14

Define Supervised Learning:

Answer 14

Agents Observe input-output pairs & learns a function that maps from input to output

Question 15

Define Unsupervised Learning:

Answer 15

Agent Learns patterns in the input without any explicit feedback

Question 16

What is Unsupervised Learning?

Answer 16

• Learning without a teacher
• Find Hidden Structures
  Clustering ==> Group inputs based on similar properties

Question 17

What is Reinforcement Learning?

Answer 17

• COMBO of Supervised & Unsupervised
• Learning with (delayed) feedback / reward –> Don’t have instant labels
• Learn series of actions => Sequence of decision Making

Agents learn from a series of Reinforcement Rewards & Punishment. Decides which of the actions prior to reinforcement were most responsible for it and alter actions towards more reward in Future.