You are on page 1of 14

Project – 1 report

On
TO PIC : Application of partial derivatives

Submitted towards partial fulfillment of

M TE COM PON EN T EVALU ATION


In
Bachelor of technology

Engineering Physics
Submitted by:

2k20/A5/65
Katyani Gupta

Under the supervision


Of
Dr. M
Jayasimhadri
DEPARTMENT OF APPLIED MATHEMATICS
DELHI TECHNOLOGICAL UNIVERSITY
BAWANA ROAD-110042
acknowledgement
▪ I would like to express my sincere
gratitude to my professor Dr. M
Jayasimhadri
▪ for his guidelines ad help while
doing this interesting project. His
guidance helped me in doing the
project, reaching to a conclusion .
▪ I would also like to express my
gratitude to department of applied
mathematics for providing such an
opportunity to think outside the
box and have such a golden
opportunity.
Brief of the project
▪ In this project, I am going to focus on usage of
partial derivatives in machine learning as we
know maths is the beauty of logic and it is behind
every subject even AI also.
▪ I have shown its application in concept of
gradient and how it is used through a simple
example of cost function .
▪ Machine learning and deep learning
applications usually deal with something called
the cost function whose minimization represents
how good or bad a machine model can be .
▪ The main reason I chose this project was to learn
about how a concept of partial derivatives is used
in machine learning algorithms and that is the
crux on which output of a machine is based.
▪ The applications I am showing are :
▪ Minimization of cost function
▪ Neural network system
▪ The research on this topic is actually claiming to
answer how in today world whatever concepts we
have learnt so far forms basic of dynamic field of
Artificial intelligence , how these models made
helps in better human understanding dealing
with strategies prepared by the very old and
basics of mathematical concepts

Total derivative vs partial


derivative
▪ The total derivative and the
partial derivative are related
but at times fundamentally
different.
▪ All constraints and variable
substitutions have to be
done before calculating the
partial derivative.
▪ The partial derivative ignores
implicit dependencies.
▪ The total derivative takes all
dependencies into account.

Concept of partial
derivatives
▪ Basically, it states that if there is a function of two
variables ,then to find partial derivatives then to
find partial derivative of that function w.r.t to one
variable , treat the other variables as constant. It
tells us the instantaneous change in the function
with respect to each variable when other is fixed ,
hence helping to draw various inferences and
conclusions
▪ Let us take an example : f(x,y)= x^4 + 4y^4
▪ Partial derivative wrt.x = df/dx =4x^3
▪ Partial derivative wrt.y = 16y^3
▪ Now this knowledge of partial derivative is used
in gradient descent to determine cost function.
And its optimization , for this new concept of
gradient descent is used.
▪ What is optimization??
▪ we always need to consider a set of values called
the feasible set or feasible region.
Mathematically speaking it is always a subset of
real numbers set X ⊆ R.
▪ Now comes the concept of partial derivatives,
This derivative describes how output changes at
point x if only variable xi changes. As you know,
when we are working on machine learning, deep
learning or AI applications

Study of the use:


▪ Let us see the example of a machine model where cost
function and its minimization is the key goal for any machine
learning algorithm as a lower error between the actual and
the predicted values signifies that the algorithm has done a
good job in learning.
▪ Cost function is of the form Y = X² but we take the account
of error and assumptions also and for complex problems , we
will use gradient descent
▪ Suppose we are on a parabola of cost function and we have to
decide as to what path we have to follow where our goal is to
reach minimum hence , two things are in our mind which path
do we have to follow and where to reach?? Gradient Descent
Algorithm helps us to make these decisions efficiently and
effectively with the use of derivatives.
▪ Let us now put all these learnings into a
mathematical formula. In the equation, y = mX+b,
‘m’ and ‘b’ are its parameters
▪ . During the training process, there will be a
small change in their values. Let that small
change be denoted by δ.
▪ The value of parameters will be updated
as m=m-δm and b=b-δb respectively. Our aim
here is to find those values of m and b in y = mx+b ,
for which the error is minimum i.e values which
minimize the cost function.
▪ To solve for the gradient, we iterate through our data
points using our new 𝑚m and 𝑏b values and compute the
partial derivatives. This new gradient tells us the slope of
our cost function at our current position (current
parameter values) and the direction we should move to
update our parameters.

Mathematical formulation
▪ this is the way where cost function
is minimized and we can easily
predict how good a machine is bad
or gold depending upon the value
incurred .
▪ Through graphical interpretation,
we can see the tangent drawn at the
path from where we are moving at
the path to find minimum value.
▪ Hence through the concept of
gradient descent with knowledge of
partial derivatives , it became
backbone of machine learning
algorithm in artificial intelligence.
A good way to make sure gradient
descent runs properly is by plotting
the cost function as the optimization
runs. Put the number of iterations on
the x-axis and the value of the cost-
function on the y-axis. This helps you
see the value of your cost function
after each iteration of gradient
descent, and provides a way to easily
spot how appropriate your learning
rate is.
Some other applications of
partial derivatives
▪ 1. utility function u=f(x,y)u=f(x,y) is a
differentiable function and a function of two
goods, xx and yy:
Marginal utility of xx, MUxMUx, is the first
order partial derivative with respect to xx
And the marginal utility of yy, MUyMUy, is
the first order partial derivative with respect to y
▪ This is we have studied in various economical
concept that is utility of how much consumption
is done by a person ,firm or an unit with respect
to certain variables , it is of great importance in
microeconomics
▪ Marginal utility is useful in explaining how
consumers make choices to get the most benefit
from their limited budgets.
▪ 2.in thermodynamics , we have studied about
all the process like isotropic
process(temperature constant), isochoric
process(volume constant), isobaric
process(pressure constant)
▪ What we are doing here keeping one parameter
constant but studying the process of
thermodynamics with one variable and deducing
expression for it , so this is what we have seen
commonly.
▪ Its usage is vast but in my project, I kept it till the
gradient descent , a basic concept in machine
learning

conclusion
▪ From the project, I have concluded the following first the
following concepts:
▪ Optimization is a big part of machine learning.
▪ Gradient descent is a simple optimization procedure that you
can use with many machine learning algorithms.
▪ Then, this concept of partial derivative is used in machine
learning alogrithms and is the crux on which the output and
performance of any machine is based.
▪ in today’s world whatever concepts we have learnt so far
forms the basic crux of dynamic field of AI leading to various
avenues such as gaming theory , how these models produced
by AI to give better human understanding deals with
strategies prepared by the very old and basics of
mathematical concepts.
▪ Partial derivatives has a scope in various fields therefore it is
required to be studied in full detail with utost concentration
especially for the technocrats who are concerned with
gaming technology.computer learning rest it is also seen in
your classics subjects formuales be it physics theories like
relativity , newton ‘s laws , in economics also which are
explained in the project
▪ Calculus and optimization is a never ending concept , it is
intitutive and requires a lot of practice and its quest to solve
problem has no bar. In my project I just started off with the very
basic concept and used in two applications but its more
interesting applications and use can be explored

Here I would like to conclude my project after discussing basic concept


of partial derivatives, its importance , one example in artificial
intelligence through machine learning, mathematical formulation. Its
other applications etc.
Remember:
”If AI is the lucrative and appealing field in this
era of technology then mathematics is the soul
and beauty behind it”

You might also like