Learning and Reasoning

ITS69404 Cognitive Computing and Applications
School of Computer Science and Engineering

Taylor’s University Lakeside Campus
Lecture 5: Learning and Reasoning

Learning Outcome
• At the end of this topic, you should be able to:
• Understand storing cases as a learning method and hence,
discuss the fundamental elements of knowledge-based AI, and
analogical reasoning that is a core process of cognition
• Aligns to Module Learning Outcome 1
Lecture 5 - Learning and Reasoning 2

Outline
• Learning by storing cases
• Nearest neighbor method
• K-nearest neighbor
• Complex cases in the real world

Introduction
• Learning by storing cases in the memory has a strong relation to
cognition
• We as cognitive agents our interaction with the world has certain
patterns of regularity
• The kinds of problems we deal with in a routine day-to-day basis are the
same in the recent past
• We do not think much as we do; memory supplies us with the answer
• The cognition architecture has three components
• reasoning, memory and learning
• In intelligence, we focus on reasoning.
• Solving problems for decision making
• Learning by storing cases shifts the balance between the components
• Learning is very important and so is memory
• We store things in the memory and the memory supplies with the answer
so that we no need to reason as much we think we need to
Learning by Storing Cases
• Looking at the shape size and orientation, the best guess would be
….. box
Visualization of Learning by storing cases
• This visualization is still abstract

• Consider a practical problem of tying shoelaces or medical
diagnoses with a set of signs and symptoms
• A case is an encapsulation of past experience
• Learning by storing cases is a powerful method that works in a very
large number of situations from tying shoelaces to medical
diagnoses
Knowledge representation
• In learning by storing cases, the idea was most similar or most
closely resembling
• How to operationalize or make it explicit?
• In terms of knowledge representation, we can represent the block
world as a two-dimensional grid
Height

Knowledge representation contd..
• The problem can be represented in the same two-dimensional grid
Height

Specific method for retrieving cases

Exercise

Exercise contd..
• The problem here is in a two-dimensional grid because two coordinates x and y

end up to any one point
• In the real-world problems are not so easy to represent
• We might need a multi-dimensional space to represent all the cases in the
problem
Multidimensional space problem

Multidimensional space problem contd..
• We can calculate the similarity of the new problem with the old
cases in terms of origin or destination



• K Nearest Neighbor (KNN) method

• KNN is a powerful method, but has limitations
• One limitation
• in the real-world no. of dimensions with which we might want to compute
the distances between the new problem and old cases could be very large
• A High dimensional space where deciding which of the stored cases is
closest to the new problem may not be simple as it here
• Second limitation
• Even if the new problem is very close to the existing case it does not mean
the existing cases solution can or should be directly applied to the new
problem
• We need alternative methods of retrieving cases from memory,
and methods for adapting past cases to fit to the requirements of
the new problem
• This is called case-based reasoning
Summary
• Recording and using cases
• is a learning method
• individual case encountered in the past is recorded to use them in the
future problem solving
• Nearest neighbor method
• Finding the most similar case to the current problem we faced in the past
• Cases in real-world problem is more difficult
• Nearest neighbor in k-dimensional problem to solve complex cases
– E.g., navigation problem with automated car
• Has limitations
• Executing a solution used in the past doesn’t works often
• Often, we have to store cases with qualitative and quantitative data
• Case-based reasoning will be able to address these problem that will
include adaptation and evaluation into the process
Q & A!

Learning and Reasoning

Uploaded by

Copyright:

Available Formats

Learning and Reasoning

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Learning and Reasoning

Uploaded by

Copyright:

Available Formats

ITS69404 Cognitive Computing and Applications

School of Computer Science and Engineering

Lecture 5: Learning and Reasoning

• Aligns to Module Learning Outcome 1

Lecture 5 - Learning and Reasoning 2

Lecture 5 - Learning and Reasoning 3

• This visualization is still abstract

Lecture 5 - Learning and Reasoning 7

• The problem can be represented in the same two-dimensional grid

Lecture 5 - Learning and Reasoning 8

Lecture 5 - Learning and Reasoning 9

Lecture 5 - Learning and Reasoning 10

• The problem here is in a two-dimensional grid because two coordinates x and y

Lecture 5 - Learning and Reasoning 12

Lecture 5 - Learning and Reasoning 13

Lecture 5 - Learning and Reasoning 14

Lecture 5 - Learning and Reasoning 15

Lecture 5 - Learning and Reasoning 16

You might also like