American Journal of Engineering Research (AJER) 2014

w w w . a j e r . o r g

Page 169
American Journal of Engineering Research (AJER)
e-ISSN: 2320-0847 p-ISSN : 2320-0936
Volume-03, Issue-05, pp-169-179
www.ajer.org

Research Paper Open Access


The Expected Number of Real Zeros of Random Polynomial

A. K. Mansingh
1
, P.K.Mishra
2

1
Department of Basic Science and Humanities, MITM, BPUT, BBSR, ODISHA, INDIA.

2
Department of Basic Science and Humanities, CET, BPUT, BBSR, ODISHA, INDIA.

Abstract:- In this paper we have estimated the number of real zeros of
Z z e =

=
, z X (Z) Q
n
0 j
j
j n
which is a random Gaussian polynomial satisfying the normal
distribution with mean zero and variance one i.e. 0 ) ( =
j
X E and 1 ) (
2
=
j
X E for j 0. Much
research works has been done on the same polynomial with different co-efficients satisfying
the above condition and found that the expected number of zeros is approximated to
( ) n log
2
t
as n in the interval (- , ). Our present work is to estimate the number of
zeros in the interval [0, 1] and found that the expected number of real zeros of the above
polynomial under same conditions is | | ( ) n log
2
1
~ 0,1 ENn as
n
. Our result gives
better approximation as compared to results given by Yoshihara [6]

Keywords: - Independent identically distributed random variables, random algebraic
polynomial, random algebraic equation, real roots.

INTRODUCTION
Let X
0
, X
1
X
n
be a stationary Gaussian process satisfying 0 ) ( =
j
X E and 1 ) (
2
=
j
X E for
j 0 as sufficient conditions. Then the expected number of real zeros of the above
polynomial

=
=
n
j
j
j n
z X z Q
0
) ( is
t 2
1
( log n ) as n tends to infinity. The expected number of
zeros of a random trigonometric polynomial has been studied by Dunnage[1] and estimated
the number of zeros in the interval [0,1] which is approximated to log ) 2 / 1 ( n t where
n but in the same problem we consider a polynomial which is piece wise continuous
and differentiable in the same interval and used normal distribution with mean zero and
variance one and applied Gaussian process. Ibragimov and Maslova [2] had worked with
same mean and variance under different conditions and had got the expected number of zeros
in the same interval is approximated to the result of Dunnage [1].
Let us consider the Gaussian polynomial

=
=
n
j
j
j n
z X z Q
0
) (
(1)
American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 170
Which is a piece-wise continuous and differentiable polynomial within a closed interval [0, 1]
and satisfying same conditions. Suppose that there is an interval [-b,b] , 0<b , where
( ) u f is uniformly approximated by the partial sums, ( ) u f
n
S of its Fourier series
development, and 0 m ( ) u f M , | | , - e where m and M are the lower and
upper bounds of ( ) f u . Then the expected number of real zeros of the above polynomial is
| | { } ( ) n log
2
1
~ 0,1 ENn as n .
Let the number of zeros of the above polynomial is denoted by EN
n.
.The main aim of our
work is to estimate | | { } 0,1 ENn .
Now we have partitioning the interval [0, 1] into three different sub intervals namely I
n
1
, I
n
2

& I
n
3
the details of partitions are given bellow,


(

=
(
(

=
(
(

=
1 ),
log log
1 (
)
log log
1 ( ),
log
1
1 (
)
log
1
1 ( , 0
3
2
1
n
n
I
n
n
n
I
n
I
n
n
n

For n3,
Let N
n
(a,b) be the number of sign changes of a piece-wise linear approximation to Q
n
(x).
First we estimate ) (
2
n
n
I EN . Then at last section we have approximated EN
n
(I
n
1
) and
EN
n
(I
n
3
) and found that the expected number of zeros of both the intervals are equivalent to
(logn) as n tends to infinity
i.e. ) (log ) (I EN ) (I EN
3
3 n
1
n n
n o = +

COVARIANCE ESTIMATES:
Let k
n
(x,y)=E{Q
n
(x)Q
n
(y)} and r
n
(x,y )=Cor {Q
n
(x) , Q
n
(y)}. We estimate k
n
(x,y) and r
n
(x,y)
for x,y satisfying certain conditions. For
In
x
2
e we derive upper and lower bounds for
k
n
(x,x).
Let T
n
) (u be a trigonometric polynomial of order n with real coefficients. Suppose we
define
| | R c and , - , e c ) (
v
n
-n v
iv
v
e e =

=
t t u u
u
n
T
(2)
( ) |
.
|

\
|
=

=
2 / x c 2 ,
0
n
0 v
v
v
c x T G
n
t (3)
Then

( )
( ) ( )
( ) ( )
(0,1] y x,
1
y - 1 x - 1
y) (x, r
(0,1] y x,
1
, ,
, ,
2
1
2
1
2 2
/
e

=
e

+
=
xy
xy
y T G x T G
y x T
n n
n
a
k
Taking s s s y) (x, r d 0
/

American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 171
We estimate k
n
(x,y) satisfying the conditions
< <d 0 And x, y
I
n
2
e
We know from co-variance estimates
have we y x, For
n as 0 ) , ( ) , (
2 1
/
0 ) , (
,
sup
2
I I n n
n
d y x r
y x
y x r y x r
I n
e

> > '
e

THEOREM -1 Suppose that ) (u f can be uniformly approximated by the partial sums of its
Fourier series development | | t t u u , , ) ( e < s A f & f (0)>0. If there is a constant and
an integer N
0
such that | | 1 , 0 x , N n for 0 ) , (
0
e > > >o x f S G
n
Applying co-variance estimates
to find mean and variance of the series
j
m
j
j
X a

=0
we have
0 m for 2
0
2
2
0
> s
|
|
.
|

\
|

= =
m
j
j
m
j
j a
j
A X a E t
and
u u
t
t
u u
)d f( ) , (
0 0
}

= =

|
.
|

\
|
|
.
|

\
|
=
n
v
iv v
n
v
iv v
n
e y e x y x k

{ } ) , ( Q(x)Q(y) E y) k(x, et y x k L
n
n
Lim

= =
( ) ( )
}

=
t
t
u u
u u d f ye xe
i i
) ( 1 1
1 1

( ) 4

=
=
m
m v
iv
v
e c
u
u) ( T Let
m
(5)
We have by Cauchys residue theorem that
( ) ( )
1
1 1
) 1 ( 2 1 1

=
}
xy x d ye xe e
v i i iv
t u
t
t
u u u

And
( ) ( )
1
1 1
) 1 ( 2 1 1

=
}
xy y d ye xe e
v i i iv
t u
t
t
u u u

( ) ( ) C d g ye xe y x h
b
b
i i
s
}

u u
u u
) ( 1 1 ) , (
1 1
(6)
Where C is a constant depending on B and b.
Taking m=n in equation (5) we have m=n and T
n
()=S
n
f(), where S
n
f() is the nth partial
sum of the Fourier series development of f(). Now for | | 1 , 0 y x, e we have
0
) , , ( ) , ( J y x f S y x k
n
a
n k
s (7)
Where
4 3 2 1 0
J J J J J + + + = and
American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 172
( ) ( ) ( )
}

=
b
b
n
i i
n
a
d f S ye xe y x f S k J u u
u u
1 1
1
1 1 ) , , (

( ) 8
( ) ( )
}

=
b
b
i i
d f ye xe y x k J u u
u u
) ( 1 1 ) , (
1 1
2
( ) 9
( ) ( )
}

=
b
b
n
i i
d f f S ye xe J u u u
u u
) ( ) ( ( 1 1
1 1
3

( ) 10

) , ( ) , (
4
y x k y x k J
n
= (11)
From the conditions of Zygmund [7] we found that there is a constant B not depending on n
such that
0. n all for B d f S
n
> < s
}

, ) (
t
t
u u & ) ( ) ( u u f S g
n
= gives < s C J
1

Where C depends on b and B & g()= f() gives
Let
| |
) ( ) ( sup ) (
,
u u
u
f f S n a
n
b b
=
e

By Cauchys inequality we have

2 / 1 2 2 / 1 2
3 ) 1 ( ) 1 (
) ( 2
y x
n a
J

s
t

) , , ( ) , ( y x f S k y x k
n
a
n



( )
( )
( ) ( )
|
|
|
.
|

\
|
+ + +
+ s

+ +
+
y
x
xy y
x
n n
n
n a
C
2
2
1
1 1
2 / 1
2 / 1
1 1
1

For | | 1 , 0 , e y x and where C is a constant depending upon b, A and B So
( ) ( )

1 1
) (
) , , ( ) , (
2 / 1
2
2 / 1
2
y x
n w
y x f S k y x k
n
a
n

s
(12)
Where n as n w 0 ) ( For e
I
n
y x
1
,
I
n
2

Now

=
+ =
n
v
v
v n
x r x f S G
1
2
1
) , (
(13)
We show that there is a constant >0 and an integer N
0
such that
0
2
n
, x 0 ) , (
I
N n when x f S G
n
> e > >o From Abels Theorem and Titchmarsh[4]
conditions we see that G(S
n
f ,x) is uniformly convergent for | | 1 , 0 x e and
( )

= + =

n
v
v
x
f r x f S G
1
2
1
1
0 ) , (
lim
t (14)

PRROF OF THEOREM -2 Using the two conditions | | t t u u , , ) ( e < s A f and f(0)>0
of Theorem-1 and applying the uniform convergence of the series within a certain interval
| | t t u u , ) ( e f S
n
and ) 0 , ( ) 0 ( f S G f S
n n
= t by adopting similar procedure as in the
proof of Theorem -1 we see that J
1
=J
2
=0 holds for
I I n n
x
2 1
e . For some we construct an
American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 173
interval | | 0 ) f( and , - > > e u t t u . Such an interval exists as ) (u f is continuous at =0 and
) 0 ( f >0. We consider the case when ( ) 1 , ) (log 1
2 / 1
e n x and x=1 separately.
The following inequality holds for | ) 1 , 0 , e y x
( ) 0 2 , b cos - 1 1
-1/2
2
1
0
sup
> > s

s s <
b xe
i
b
t
u
t u
2 for 1 1
1
0
sup
t t
u
t u
> > s

s s <
b xe
i
b
We
have
| ) 0,1 x for
0
0
sup
e

=
s s <

u
t u
iv
n
v
v
b
e x

Substitution in k
n
(x,x) with by replaced 1
2
u iv
xe

2
0
u iv
n
v
v
e x

=

We got
| ) 0,1 x C ) ( - x) (x, k
2
b
b -
n
0 v
n
e s
}

=

u u
u
d f e x
iv v
(15)
Substitute t u 2 / 1 ) ( = f in (15) we have
C
n
0
n
0
b -
2
-
2
s
=

=

}

}

b
d
iv v
d
iv v
e x e x
u
v
u
u
v
u
t
t
(16)

Where C depends only on b and A.

By simple calculation we have
| | 1 , 0 x , 2
n
0 v
2
2
-
n
0 v
e =

}

= =
v iv v
x d e x t u
t
t
u
(17)

}

= =

n , 2 ~
n
0 v
2
2
b -
n
0 v
v
b
iv v
x d e x t u
u
(18)

and
| | 1 , ) (log 1
2 / 1
e n x
.

From our construction of [-b,b] we have
u u u
u u
d e x d f e x
b
iv v
b
iv v
2
b -
n
0 v
2
b -
n
0 v
) (
}

}

=

> (19)
So the desired result follows for | ) 1 , ) (log 1
2 / 1
e n x .
When x=1 we have
( ) ( ) ( ) ( ) 0 1 2 1 2 1 1 , 1
1
f n r j n n
n
n
j
j
n k o
t

=
+ = + + + =
(20)
American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 174
Where
n
f() is the nth Cesaro sum associated with f(). As f() is continuous at =0 we
have
n
f(0) n as f ) 0 ( . Hence we
have u u
t
t
u u
d f e a e a x a E
v
iv
v
v
iv
v j j
) (
-
m
0
m
0
2
m
0 j
}

|
.
|

\
|
|
.
|

\
|
=
|
|
.
|

\
|
= =

=

given that . ) ( < s A f u Then by using Cauchys inequality we have
u
t
t
u
d e a A x a E
iv
v j j
}

=

=
s
|
|
.
|

\
|
-
2
m
0 v
2
m
0 j
(21)

By simple calculation we have

= =

=
m
j
j
iv
v
a d e a
0
2
-
2
m
0 v
2t u
t
t
u
(22)

GENERAL APPROXIMATION FORMULA : - For | | b a x , e we approximate Q
n
(x) by
a process which linearly interpolates between Q
n
(a) and Q
n
(b). It is convenient to count the
sign changes of this process in [a,b] by
{ } (b) (a)Q Q sgn
2
1
2
1
) , (
n n
|
.
|

\
|

|
.
|

\
|
= b a N
n
(23)

Where sgn{x} =

<
=
0 x 1 -
0 x 0
0 > x 1

The results of this section depend on an upper bound for the number of zeros of Q
n
(x) in an
interval [a,b]. Define the event

b] [a, in zeros more k has ) (x Q U
n k
=
(24)
For any interval [a,b]
] 1 , 0 [ _
let
| |
| |, 1 , 1) (n - 1 b ), )( 1 (
and , 1) (n - 0,1 b , ) 1 )( (
1 -
-1 1
+ e + =
+ e =

a b n
b a b

LEMMA -1 let and
30
2

<
(i) ) (u f is continuous at u =0
(ii) 0 ) 0 ( > f
(iii) | | t t u u , - , ) ( e < s A f
Then there is an integer N
1
and an absolute constant C such that
0 k N n ) (
1
5 / 3
> > <
k
k
C U P
COROLLARY-1 Under the conditions of Lemma-1 we have
N n , ]) , ([ ) , (
1
5 / 6
> < C b a EN b a EN
n n

Where C is an absolute constant.
Now we have to estimate the number of zeros in the three sub intervals.
(A). ESTIMATION OF NUMBER OF ZEROS IN THE INTERVAL:-
I n
1


American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 175
LEMMA -2 Let X
0
, X
1
X
n
be a set of n points satisfying 0 ) ( =
j
X E and 1 ) (
2
=
j
X E for
j 0 .
(i)Let us consider a function ) f( which is uniformly approximated by the partial sums,
S
n
) f( in the interval [-,]
(ii) In the interval 0 m ) f( M and | | , - e where m and M are the lower
and upper bounds of ) f( . Then expected number of real zeros in the interval
(
(

= )
log
1
1 ( , 0
1
n
I n is ( ) 2 n n log log 2 / ) (
2 / 1
1
> s t C ENn
In
(25)
Where C is a finite constant
PROOF: - From the Kac-Rice [3] formula and using the postulates of Shankar [4] which
states that
| | ( ) dx b a EN
R B C n
b
a
n n
n
2
2 2 / 1 2 / 1
1 )
1
( ) , ( =
}

t
(26)
( ) ) ( E C , )) ( (
2
/
n
2
x Q x Q E B n
n n
= = (27)
( ) ) ( ), (
/
x Q x Q Cor R n
n n
= (28)
Where ) (
dx
d
x Q
n
= ) (
/
x Qn
Now we have to find an upper bound for
I
1
n
with x / e
n n
B C
We represent B
n
=k
n
(x,x) by Noting that | | t t u u , , 0 ) ( e > > m f and applying
Lemma-1 gives 2
0
2

=
>
n
v
v
n
x m B t for | | 1,0 x e From Lemma-2 we have
( )
|
.
|

\
|
s

=

n
1
1 2 2
2
v
iv
v
v
n
e a x v A C
u
t summing

=
n
1
2
v
v
x and using that ) 1 ( 2
3 2 ) 1 ( 2
1
2
=
s

x x v
v
n
v

| ) 0,1 x e (29)
( ) x - 1 ) 1 )( / 2 ( /
-2
2 1 ) 1 ( 2 +
s
n
n n
x m A B C

(30)
2 n for x , 1
I
1
n
) 1 ( 2
> e <
+ n
x
e
+
e
n as x
n
I
1
n
) 1 ( 2
x
x , 0
I
sup and
1
n
We deduce that
( )
( )
1
1 n 2
x - 1

+
is
bounded above by a constant. So
( )
I
1
n
-2
x and 2 n x - 1 / e > s
n n
B C (31)
Substituting the value of C
n
/B
n
and using the relation 1 ) 1 (
2
s
R
n
gives the expected
number of zeros in the interval
I
1
n
is ( ) 2 n for n log log 2 / ) (
2 / 1
1
> s t C ENn
In

(32)
Hence the theorem is proved.
LEMMA-3 Let X
0
, X
1
X
n
be a set of n stationary points satisfying 0 ) ( =
j
X E and
1 ) (
2
=
j
X E . for j 0 Suppose that (i) there is an interval | | t t, and there is a constant
American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 176
such that

=
> > +
n
1 j
j
j
0 x r
2
1
o | | , 1 , 0
0
N n x > e for some integer N
0
there is a
constant C such that
( )
5
2 / 1
1
N n for logn log
2
>
|
|
.
|

\
|
s
t
C
EN
In n

PROOF: - To find an upper bound for C
n
/B
n
, x
I
n
1
e and applying Lemma-1 and Lemma-2
we have
( )
0
1
N n for 0
2
1
, > > > + =

=
n
j
j
j n
x r x f S G o
(33)
Let
n
N be any integer
So the conditions of Theorem-2 are satisfied. From Theorem-2 and any ( ) 1 , 0 e C we
have an integer N
4
such that
x & N n ), , , ( ) , (
I
1
n
4
e > > = x x f S CK x x k B
n
a
n n

(34)
Where C being any constant and (0,1) Ce
From LEMMA-2 we have
2
1
) , (
2 ) , , (
x
x f S G
x x f S K
n
n
a

=
But
0 n
N n for x f S G > > > 0 ) , ( o . So
) ,N (N max N n ,
1
2
4 0 5
2
= >

>
x
C
B
n
o
(35)
In the same way as in Lemma- 2 we have
( ) . ) ,N (N max n , x - 1 /
4 0 5
-2
= > s N C B
n n
Then the expected number of zeros in the
interval
I
3
n
is
( )
5
2 / 1
1
N n for logn log
2
>
|
|
.
|

\
|
s
t
C
EN
I n n
(36)
Hence the theorem is proved.
LEMMA-4 Let ( ) 0 , > j Z
j
be a stationary sequence of uniformly mixing random variables
with zero mean and
(i) 0 for
2
> <
+
o
o
j
Z E
(ii).
|
|
.
|

\
|

=
n as
2
0
n
j
j
Z E
Then there exists a constant C such that

2 / 1
2
0
2
0
o
o
+
=
+
=
|
|
.
|

\
|
s

n
j
j
n
j
j
Z E C Z E

THEOREM- 2 Let X
0
, X
1
X
n
be a set of n stationary real-valued uniformly
mixing Gaussian process with 0 ) ( =
j
X E and 1 ) (
2
=
j
X E for j 0 satisfying (i)
<

1
) (
2 / 1
k
k | and (ii) ) f( >0 and also satisfied by the sequence { } 0 j , X (-1)
j
j
> .
American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 177
We have
( ) ( ) logn log n C EN
1/2
n
n
I
log
1
s
Where C is a constant.

PROOF: Choose C such that
( ) 1
0
< = > q c X P and Choose the events ( ) c X B < =
0 0
and
( ) 1,....n. k and , ......
1 0
= > < < =

c X c X c X B
k k k

( ) c X c X B
n
< < = ......
0

Let R r < < 0 .
Using the argument of Ibragimov and Maslova[2]
We obtain.
| | ( )

}
= =

+ + s
n
k
n
k
B
s k n
k
dP H r R B nP B kP r r EN
1 0
1
) / (log ) ( ) ( , Where
| |
|
.
|

\
|
=

e
) (R ) ! ( sup log
) (
1
, 0
u
t t
i
e
k
n
s
Q c k H
(37)
Here Q
n
(k)
(x) is the k
th
derivative of Q
n
(x) with respect to x.
We estimate P(B
k
) and P(B)
Define a sequence of random variables
(Z
k
, k=0,.n) by

X q -
X 1
k
k

<
>
=
c
c q
Z
k

Now for k=1..n we have
( )
) 1 (
1
0 0
1
0
q Z B
and q Z kq Z B
k
k
j
j k
= =
|
|
.
|

\
|
=
|
|
.
|

\
|
= =

=
=

We show that the conditions of Lemma-4 are satisfied by (Z
k
, k0) with 4 = o Clearly (Z
k
,
k0) is a stationary sequence of uniformly random variables with mixing coefficient ) ( j | .
Using Ibragimov [2] and

=
<
1
2 / 1
) (
j
j | (38)
gives

|
|
.
|

\
|

=
n as n Z E
n
j
j
~
2
1
0
(39)

So condition (i) & (ii) is satisfied. Using, Markovs inequality and Lemma-4 with
4 = o gives
1 / 2 / ) (
3
0
> < <
|
|
.
|

\
|
> s

=
k and C f or k C kq Z P B P
n
j
j k
And <

=
) (
1
k
k
B kP

American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 178
In the same way
< <
|
|
.
|

\
|
+ > s

=
C n C n q Z P B P
n
j
j
, / 2 / ) 1 ( ) (
3
0
(40)
And n , 0 ) (B nP
We estimate the final term in using the method of Ibragimov and Maslova [2]
) exp( ) 1 ( log ) ( log ) (
0 0
i i C T B P w B P dP H
k k k
B
s
k
+ + + s
}
(41)
Where C is a constant, ) ln(
0
T i = and T>0 Then
|
|
.
|

\
|
+
=

=

j
n
k j
k j
k
X
c k
R k j j j
E W
!
) 1 )....( 1 (
Taking T to be the following function of k

>
=
+
1 k ,
0 = k 1,
1 c
k
T

0 for > c
using Kac-rice formula and noting that
1 0 ) 1 (
1 1
< s <

R c R C W
k
k
we have

}
= =
|
.
|

\
|
+ |
.
|

\
|

s
n
k
k
n
k
B
s
B P k
R
C
dP H
k
0 0
) ( ) 1 (
1
log
) ( ) (log ) 1 (
0
k
n
k
B P k

=
+ + c

=
+

+ +
+
n
k
c
k
k
D
1
1
log
log ) 1 ( 1

c
c



On Substituting r = 1- (logn)
-1/2
and R=1-1/2 (logn)
-1/2
in the above equation we get the
expected number of real zeros in the interval
1
n I is ( ) 2 n for n log log 2 /
2 / 1
> t C where C
and D are constants

(B) ESTIMATION OF NUMBER OF ZEROS IN THE INTERVAL:- ( )
I n
2

To find out the expected number of zeros in the interval
number
(
(

= )
log log
1 ( ),
log
1
1 (
2
n
n
n
I n
Let the expected number of zeros of the above interval is denoted by ( )
I
n
n
EN
2

LEMMA -5 Suppose that t s <b 0 If
(i) ) (u f can be uniformly approximated in b] [-b, by the partial sums of its Fourier
series development, ) (u f S
n

(ii) f(0)>0
(iii) | | t t u u , - , ) ( e < s A f
Then ( ) n log ) 2 / 1 ( ~
2
as n EN
In n
t




American Journal of Engineering Research (AJER) 2014


w w w . a j e r . o r g

Page 179
(C) ESTIMATION OF NUMBER OF ZEROS IN THE INTERVAL ) (
3
In
:-
LEMMA -6 If
(i) ) (u f is continuous at 0 = u
(ii) 0 ) ( > u f
(iii) | | t t u u , - , ) ( e < s A f
Then there is a constant and an integer N
6
such that
( )
6
5 / 7
3
N n , ) log (log > < n C EN
In n

THEOREM -3 Let X
0
, X
1
X
n
be a set of n points satisfying 0 ) ( =
j
X E and 1 ) (
2
=
j
X E .
For j 0. Suppose that there is an interval [-b, b], 0<b , where ) f( is continuous at =0,
| |, , - e Then expected number of real zeros in the interval
(

= 1 ),
log log
1 (
3
n
n
I n is
( )
6
5 / 7
3
N n ) log (log > < n C EN
In n

Where C is a finite constant
PROOF : - Now we have to find an upper bound for
I
3
n
with x / e
n n
B C using Kac-Rice
formula ( ) x - 1 ) 1 )( / 2 ( /
-2
2 1 ) 1 ( 2 +
s
n
n n
x m A B C So
( )
I
3
n
-2
x and 2 n x - 1 / e > s
n n
B C
Substituting for C
n
/B
n
in the above equation and n taking the help of the inequality
1 ) 1 (
2
s
R
n
the expected number of zeros in the interval
I
3
n
is
( )
6
5 / 7
3
N n ) log (log > < n C EN
In n
.
Hence the theorem is proved.

REFERENCES
[1] Dunnage, J.E.A. The number of real zeros of a class of random algebraic polynomials
(I), Proc. London Math. Soc. (3) 18 (1968), 439-460.
[2] Ibragimov, I.A. and Maslova, N.B. The average number of real roots of random
polynomials, Akad. Nauk. SSSR. (199), (1971). 13-16.
[3] Kac, M. Rice, S.O. On the average number of real roots of random algebraic equation
(II), proc. Londan. Math. Soc.(1949), 390-408.
[4] Shenker, M. The mean number of real zeros of one class of random polynomials,
Annals of Prob. 9 (1981), 510-512.
[5] Titchmarsh, E.C. Theory of Functions, 2nd edn, The English language Book Society,
Oxford University Press, 1939.
[6] Yoshihara, G. (1979). The average number of real zeros of a random algebraic
equation. Bull.Amer. Math. Soc. 54, 125-134.
[7] Zygmund, V. (1969) On the average number of real roots of a random algebraic
efunctions. . Proc. London Math. Soc. 11, 238-345.