Parity function definition and intuition, characteristic function of a set.

928 Views Asked by At

I have a question of two (and a half) parts relating to parity functions. I) Pertaining to the definition of a parity function. II) Pertaining to the intuition behind checking some specific functions' parity.

I) In this particular set of questions, the parity function is defined as (image link here): Parity function $Pf(X)$ is equal to $1$ if an even number of inputs is $1$ (i.e the function being checked, further on $f(x)$, is even), and $-1$ otherwise.

This is not in line with the typical "Parity Bit" function which, unlike this one, returns "$1$" if the $f(x)$ is odd. I was wondering whether my case was a specific definition or a distinct widely known separate one.

II) I would like my intuition checked on the following exercise, if possible, explained why/where I was wrong.

For which of these does the parity function return $1$ for? In other words, which of these are "even"? The dimension of input is $n>1$.

1) Function that returns the least-significant (right-most) bit of the argument.

For a two-bit variable there's only one possible input where it returns a "$1$". Thus, this one is odd.

2) Function that returns k-numbered bit of argument, where k is less than n.

Same logic as previous case.

3) const f(x) = 1

There's $2^{(n-1)}$ possible inputs for every $n>2$, and there's an equal amount of possible outputs. The function is even.

4) const f(x) = 0

If you consider 0 an even number, then the logic from before applies.

5) "Characteristic function of a set with the cardinality of 5"

This seems to be referring to an indicator function. My understanding is limited, but it wouldn't seem like this is even.

6) f(x)=1 only if x has an odd number of ones when written in binary

For every n, half of the inputs have an odd number of ones, which is $2^{(n-2)}$. For $n=2$ this number is equal to $1$, so it's incorrect to generalize this function's evenness.

7) f(x)=1 only if x has an even number of ones when written in binary

For every n, half of the inputs have an even number of ones, which is $2^{(n-2)}$. For $n=2$ this number is equal to $1$, so it's incorrect to generalize this function's evenness.

1

There are 1 best solutions below

0
On

We define $\tilde x\in\mathbb{Z}$ and $\tilde x_i\in\mathbb{Z_2}$ (is a bit-value),

and $\tilde x_i = 1-2x_i$ is clearly represented like the following function:

\begin{align*} f(x) = \begin{cases} 1 &\text{if $x$ = 0}\\ -1 &\text{if $x$ = 1} \end{cases}\end{align*}

When $x$ (the input bit) is equal to $0$ the function returns $1$, else if $x$ is equal to $1$ the function returns $-1$.

The length of the bitstream of $\tilde x$ can be calculated this way: $$N = \lfloor\log_2{\tilde x}\rfloor+1$$

The parity function is defined as \begin{align*} P(\tilde x)=\prod_{i=0}^{N-1}f(\tilde x_i), \end{align*}

The capital pi symbol represents multiplication of successive values. Hence, we multiply together all the parities. And since $f(x)$ returns either $1$ or $-1$ then $P(\tilde x)$ will return either the value $-1$ or the value $1$. The proccess of $P$ is iterative, and the value $-1$ is never changed unless a successive parity is also $-1$ (then it changes its sign), this parity will flip as many times as there are one's in the binary expansion (of $\tilde x$) and keep the final state when $i=N-1$. The final state depends on the number of $1$'s in the binary expansion basically. In short, everytime $\tilde x_i=1$ the product will change its sign, but when $\tilde x_i=0$ the product will do nothing to the result. These two series of actions can easily be understood when $x\cdot -1=-x$ and $x\cdot 1=x$. i.e.:

$$+\cdot+ = +$$ $$+\cdot- = -$$ $$-\cdot+ = -$$ $$-\cdot- = +$$

From the above reasonings we can deduce some facts:

  1. when $\tilde x$ has an odd number of $1$'s in its binary expansion, $P$ will return $-1$.
  2. when $\tilde x$ has an even number of $1$'s in its binary expansion, $P$ will return $1$.

Examples:

$$P(2_{10}) = P(10_2) = -1$$ $$P(13_{10}) = P(1101_2) = -1$$ $$P(27_{10}) = P(11011_2) = 1$$ $$P(60_{10}) = P(111100_2) = 1$$

Other facts:

The bitlength itself do not influence the parity.


Answers:

1) Function that returns the least-significant (right-most) bit of the argument.

$$P(00_2) = 1$$ $$P(01_2) = -1$$ $$P(10_2) = -1$$ $$P(11_2) = 1$$

There are two possible inputs where it returns a $1$: $0$ and $3_{10}$.

2) Function that returns k-numbered bit of argument, where k is less than n.

Im not sure what you mean by k-numbered bit of argument. Assuming you mean the bitlength of the argument, this does not influence the output parity.

3) const f(x) = 1

The function is only even when $\tilde x = 0$ or there is an even number of $1$'s in the binary expansion of $\tilde x$.

4) const f(x) = 0

$0=$ parity $1$

5) "Characteristic function of a set with the cardinality of 5"

Assuming you mean the parity $1$ is even, then $5$ is even.

6) f(x)=1 only if x has an odd number of ones when written in binary

False, $P(\tilde x)=-1$ if there is an odd number of ones.

7) f(x)=1 only if x has an even number of ones when written in binary

True, $P(\tilde x)=1$ if there is an even number of ones.