ASANSOL ENGINEERING COLLEGE

DEPT:COMPUTER SCIENCE AND ENGINEERING

Name: TANVEER HUSSAIN
Roll no: 10800120035 Batch: A
Subject: Pattern Recognition
Code : PEC-IT602D

1. Minimum distance classifiers using Euclidian and normalized

Euclidian distance.
Minimum distance classifiers are a family of simple and widely used classifiers in
machine learning. They classify a new observation by finding the class whose
average distance to the observation is the smallest among all classes. Two
commonly used distance metrics for minimum distance classifiers are Euclidean
distance and normalized Euclidean distance.
Euclidean distance is the most common distance metric used in machine learning.
It is defined as the straight-line distance between two points in Euclidean space,
and can be calculated using the formula:
d(x, y) = sqrt((x1 - y1)^2 + (x2 - y2)^2 + ... + (xn - yn)^2)
where x and y are two n-dimensional vectors representing two observations, and
xi and yi are the ith components of the vectors.
To classify a new observation using the minimum distance classifier with
Euclidean distance, we calculate the distance between the new observation and
the mean of each class, and then assign the new observation to the class whose
mean is closest to it.
Normalized Euclidean distance is a variation of Euclidean distance that scales each
feature by its standard deviation. This metric is useful when the features have
different units or scales, and can be calculated using the formula:
d(x, y) = sqrt(((x1 - y1)/s1)^2 + ((x2 - y2)/s2)^2 + ... + ((xn - yn)/sn)^2)
where s1, s2, ..., sn are the standard deviations of the features.
To classify a new observation using the minimum distance classifier with
normalized Euclidean distance, we first normalize the features of the new
observation using the means and standard deviations of the training data, and
then calculate the distance between the normalized observation and the mean of
each class. We assign the new observation to the class whose mean is closest to
the normalized observation.

2. Show that the dual SVM problem is

Symbols used have their usual meaning.
3. Application of SVM in remote sensing image classification
Support Vector Machines (SVMs) are widely used in remote sensing image
classification tasks due to their ability to handle high-dimensional data and
their ability to perform well with limited training data. Here are some of the
applications of SVM in remote sensing image classification:
• Land use and land cover classification: SVMs can be used to classify
remote sensing images into different land use and land cover categories.
SVMs are able to capture the non-linear relationships between the
spectral information and the land cover classes.
• Object detection and segmentation: SVMs can be used to detect and
segment objects of interest in remote sensing images, such as buildings,
roads, and vegetation. SVMs can be trained on a set of features extracted
from the images, such as texture, shape, and spectral information.
• Change detection: SVMs can be used to detect changes in remote sensing
images over time. SVMs can be trained on a set of features extracted from
the images, such as spectral indices, texture, and spatial information.
• Hyperspectral image classification: SVMs can be used to classify
hyperspectral remote sensing images, which contain hundreds of spectral
bands. SVMs can be trained on a subset of the spectral bands, or on
features extracted from the spectral bands, such as principal component
analysis (PCA) or linear discriminant analysis (LDA).
• Crop classification: SVMs can be used to classify crops in remote sensing
images, which can be useful for monitoring crop health and predicting
yields. SVMs can be trained on a set of features extracted from the
images, such as vegetation indices, texture, and spatial information.
4. The equations of population and sample variance are as follows:

Justify why the denominator of sample variance is (n-1) while that of

population variance is (n) [or N]

The variance is a measure of the spread or dispersion of a set of data points.

It tells us how far the data points are from their mean value.
The equation for population variance is:
σ² = Σ(xi - μ)² / N
where:
σ² is the population variance
Σ is the sum of the squared deviations from the
mean xi is the value of the ith data point
μ is the population mean
N is the total number of data points in the population
The equation for sample variance is:
s² = Σ(xi - x)²̄ / (n - 1)
where:
s² is the sample variance
Σ is the sum of the squared deviations from the sample
mean xi is the value of the ith data point
x̄is the sample mean
n is the total number of data points in the sample
The denominator of the sample variance formula is (n-1) instead of n because
when we calculate the variance of a sample, we are estimating the population
variance based on a subset of the data. The sample mean is used to estimate the
population mean, which means that we lose one degree of freedom in the
process. Degrees of freedom refer to the number of independent pieces of
information available to estimate a parameter. In this case, we have n data points,
but we use the sample mean to estimate the population mean, which means we
only have n-1 independent pieces of information left to estimate the variance.
Dividing by (n-1) instead of n corrects for this loss of degrees of freedom and
produces an unbiased estimate of the population variance.
On the other hand, the denominator of the population variance formula is N,
which is the total number of data points in the population. In this case, we have
access to all the data points in the population, so we can use the population
mean to calculate the variance without losing any degrees of freedom. Therefore,
we divide by N in the population variance formula.