Scribd
Upload
8 views

Module 1 Part A

Uploaded by

Niveditha.p
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
8 views

Module 1 Part A

Uploaded by

Niveditha.p
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

Module 1: Introduction

May 7, 2024
● What is Machine Learning (ML)

● Example of Pattern Recognition


System
Topics ● The Design Cycle

● Learning and Adoption


Objective

Learn the overview of Machine perception, Machine Learning examples and


applications.
What is Machine Learning?
How we generate and consume data?
Can a car drive without a
driver?

Can a robot do heart


operation?
Can a machine identify object?

Can a machine identify your


voice?
Data MIning:Machine Machine learning helps
learning methods us find solutions to Algorithm for reducing
applied to large many problems in vision, the error on a set of
databases is called speech recognition,and training data.
data mining. robotics.

1 2 3 4 5

Machine learning system is Program to optimize a


also a part of artificial performance criterion
intelligence. that is in a using training data or
changing environment system past experience.
should have the ability to learn.
Scenario #1 and Example

Suppose that a fish packing plant wants to automate the


process of sorting incoming fish on a conveyor belt
according to species. As a pilot project it is decided to try
to separate sea bass from salmon using optical sensing.
Scenario #1 and Example

Salmon Fish Sea boss fish

● Sea bass is generally longer than a Salmon (Length).


● Using design or training samples of the different fish, (somehow) make
length measurements and identify ,critical value of length I*.
● If length of a fish > critical value I* then classify it as Sea Boss
otherwise it is Salmon.
● Inspect the results using histograms of two fish categories using
length features.
Scenario #1 and Example

The objects to be classified are first sensed by a transducer


(camera),whose signals are preprocessed, then the features
extracted and finally the classification emitted (here either
“salmon” or “sea bass”). The prototype of this particular task
of classification is as shown in the next figure.
Scenario #1 and Example

● Set up a camera.
● Identify physical differences like
length, lightness, width, number and
shape of fins, position of mouth etc as
a features.
● Using these features for classifier to
identify and classify the two fishes
(Object)
Pre Processing: First the camera captures an image of the fish. Next,
the camera’s pre- signals are preprocessed to simplify subsequent
operations without losing relevant information.
Segmentation: In particular, we might use a segmentation operation in
which the images of different fish are somehow isolated from one
another and from the background.
Feature Extraction: The information from a single fish is then sent to a
feature extractor, whose purpose is to feature extraction reduce the
data by measuring certain “features” or “properties.”
Classifier: These features (or, more precisely, the values of these
features) are then passed to a classifier that evaluates the evidence
presented and makes a final decision as to the species.

Pattern Classification Model


Inspect the results using histograms of two fish categories using length
features.
Key observations of length based fish classifier.

● No single threshold value l∗ (decision boundary) will serve to


unambiguously discriminate between the two categories.
● Length alone is not sufficient to classify the fishes.
● The value l∗ marked will lead to the smallest number of errors, on
average.
● sea bass are somewhat longer than salmon, on average, but it is
clear that this single criterion is quite poor; no matter how we
choose l*, we cannot reliably separate sea bass from salmon by
length alone.
Alternative feature for fish classifier.

● we try another feature the average lightness of the fish


scales.
● The Histograms for the lightness feature for the two
categories is as shown in below figure.
Alternative feature for fish classifier.
Key observations of lightness based fish classifier.

● No single threshold value x* (decision boundary) will serve to


unambiguously discriminate between the two categories; using
lightness alone.
● The value x* marked will lead to the smallest number of errors,
on average.
● The results are much more satisfactory the classes are much
better separated.
Cost in lightness based fish classifier.

For instance, as a fish packing company we may know that our customers
easily accept occasional pieces of tasty salmon in their cans labeled “sea
bass,” but they object vigorously if a piece of sea bass appears in their cans
labeled “salmon.”

Deciding the fish was a sea bass when in fact it was a salmon is Costly.To
avoid this consequence, we adjust our decision boundary to avoid cost, even if
it means that more salmon makes its way into the cans of sea bass.
What is our true task?

our true task is to make a decision rule (i.e., set a decision boundary) so as
to minimize such a cost.

To improve recognition, then, we must resort to the use of more than one
feature at a time. Now we have two features for classifying fish — the lightness
x1 and the width x2.

Thus the feature vector reduced to two dimension ie a space and we need
find a decision boundary to segregate two categories as follows,
We realize that the feature extractor has thus reduced the image of each fish to
a point or feature vector x in a two-dimensional feature space.
Overall classification error on the data shown is lower than if we use only one
feature but there will still be some errors.

Key observations on two dimension feature based fish classifier


If our models were extremely complicated, our classifier would have a decision
boundary more complex than the simple straight line. With such a “solution,”
though, our satisfaction would be premature because the central aim of
designing a classifier is to suggest actions when presented with novel
patterns(?), i.e. fish not yet seen. This is the issue of generalization. The
complex decision boundary always provides good generalization.

Key observations on two dimension feature based fish classifier


● Even with a vast amount of
training data in a continuous
feature space probability
distribution is same.
● Belief that Underfit and Overfit
model’s performance are not
Tradeoff satisfactory.

● Obtaining a better performance.


● with “simplified” recognizer,
Conclusion:

The aim of the system is to,

● Make decisions/decision Boundaries are fundamentally task or cost


specific.
● Create a single general purpose artificial pattern recognition device.
● Perform statistical pattern recognition with the statistical properties of the
patterns.
● Handle the noise associated with data.

You might also like