Scribd
Upload
17 views

Data Science Lecture

This document discusses the essential mathematics concepts needed for data science, including linear regression, principal component analysis, singular value decomposition, and linear classification. It covers topics such as dimensionality reduction, eigendecomposition, and using linear algebra principles for machine learning algorithms.

Uploaded by

Rupal Pursharthi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
17 views

Data Science Lecture

This document discusses the essential mathematics concepts needed for data science, including linear regression, principal component analysis, singular value decomposition, and linear classification. It covers topics such as dimensionality reduction, eigendecomposition, and using linear algebra principles for machine learning algorithms.

Uploaded by

Rupal Pursharthi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 24

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/359730625

Essentials of Mathematics in Data Science

Presentation · April 2022

CITATIONS READS

0 19,675

1 author:

R. C. Mittal
Jaypee Institute of Information Technology
164 PUBLICATIONS 4,064 CITATIONS

SEE PROFILE

All content following this page was uploaded by R. C. Mittal on 05 April 2022.

The user has requested enhancement of the downloaded file.


Essentials of Mathematics in Data Science

R.C. Mittal
(Formerly Prof. IIT Roorkee)
Department of Mathematics
Jaypee Institute of Information Technology
Sector 62 NOIDA (U.P.)
Outlines
• Mathematics Needed in Data Science
• Linear Regression
• Principal Component Analysis
• Singular Value Decomposition
• Power Method
• Linear Classification
Need of Mathematics
• Differential Calculus
• Statistics and Sampling Theory
• Optimization
• Linear Algebra
- Dimension
- Eigen Values and Eigen Vectors
- Inner Product and Norms
• Soft Technologies
- Neural Network
- Support Vector Machine
Singular Value Decomposition
Suppose A is a real mxn matrix, then singular
values of A are nothing but square roots of
non-negative eigen values of ATA . Here ATA is a
self adjoint matrix.
Example- Consider the matrix

A= (1)

So AT = and ATA = (2)


R. C. Mittal Applications of Linear
8
Algebra
Eigen values of ATA are (59 +/- 56.009)/2 i.e.
57.504 and 1. 495.
Therefore singular values of A are = 7.583 and
1.223 .
In singular value decomposition of the matrix A,
we decompose the matrix A into UDVT , U is a
mxm orthonormal matrix, V is a nxn
orthonormal matrix and D is a mxn matrix
with rxr diagonal matrix containing singular
values of A.
Here r is the rank of the matrix A.
R. C. Mittal Applications of Linear
9
Algebra
Principal Component Analysis
Today a lot of data is generated daily in different
social sites. In fact 90% of today data is
generated in the last 3-4 years. This data is to be
properly analyzed to gather important and
relevant information.
PCA is a dimension reduction technique which
give most dominant features from the data.
These newly extracted features are known as
“Principal Components” .
Key Points of PCA
• A principal component is a linear combination of
the original variables
• Principal components are extracted in such a way
that the first principal component explains
maximum variance in the dataset
• Second principal component tries to explain the
remaining variance in the dataset and is
uncorrelated to the first principal component
• Third principal component tries to explain the
variance which is not explained by the first two
principal components and so on
It may be noted that each additional dimension
we add to the PCA technique captures less and
less of the variance in the model. The first
component is the most important one,
followed by the second, then the third, and so
on.
How to find Principal Components?
Following steps are used in PCA
(1) Normalization of Data – Compute the mean

R. C. Mittal Applications of Linear


12
Algebra
Now normalize the data in [0,1] by using formula
z=

2. Computation Covariance Matrix- The aim of


this step is to understand how the variables of
the input data set are varying from the mean
with respect to each other, or in other words, to
see if there is any relationship between them.
Because sometimes, variables are highly
correlated in such a way that they contain
redundant information.
So, in order to identify these correlations, we
compute the covariance matrix

(4)

This is a symmetric matrix.


Cov(x, x) = var(x) etc.

R. C. Mittal Applications of Linear


14
Algebra
3. Computation of Eigen values and Eigen Vectors-
Now compute the eigen values and corresponding eigen
vectors of the covariance matrix (4). You will find some eigen
values are very small in magnitude . Precisely these eigen
values can be neglected. Thus reducing the dimension.

Principal components are new variables that are constructed


as linear combinations or mixtures of the initial variables.
These combinations are done in such a way that the new
variables (i.e., principal components) are uncorrelated and
most of the information within the initial variables is
squeezed or compressed into the first components.

So, the idea is 10-dimensional data gives you 10 principal


components, but PCA tries to put maximum possible
information in the first component, then maximum
remaining information in the second and so on.

R. C. Mittal Applications of Linear


15
Algebra
Classification
Linear Classifier
Unlike in linear regression where we predict a
numerical value, here we predict a class such
as winner or loser; rich or poor etc.
This is one of the important problem in data
analysis. We may predict
(i) A team will win or not ?
(ii) An individual will like a movie or not ?
Our input here is a point X Rn , where each
element x in X also has an associated label y
having value -1 or 1.
Suppose X = (x1 , x2 , ……, xn ) is a point set and
W = (w1 , w2 , ……, wn ) is a weight vector. We wish
to design a linear predictor as follows
y = b + w1x1 + w2x2 + …….. + wnxn
View publication stats

References
1. J.M. Phillips “ Mathematical Foundations for
Data Analysis “ (2018)
2. R. C. Mittal “Some Applications of Linear
Algebra in Computer Science”
https://www.researchgate.net/publication/3
52180145_Some_ApplicAtionS_of_lineAr_Alg
ebrA_in_computer_Science
3. G. Strang , “Introduction to Linear Algebra”
MIT Publication (2016)

You might also like