You are on page 1of 8

26/6/2021

Brief explanation of how I


prove NP≠P
Sing Kuang Tan
singkuangtan@gmail.com

Link to my paper
Prove Np not equal P using Markov Random Field and
Boolean Algebra Simplification
https://vixra.org/abs/2105.0181

Example of an Non-Deterministic problem (NP)


expressed in Markov Random Field representation
a1 a2
Example of an NP problem expressed as a Graphical model (or
Markov Random Field)

a3 a4

𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎 = ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )


, , ,

Each variable ai can take finite values, e.g. 𝑎 ∈ {0,1,2,3}

ℎ 𝑎 ,𝑎 = 0 𝑜𝑟 1

If the NP problem has a solution, 𝑓 𝑎 , 𝑎 , 𝑎 , 𝑎 >0


26/6/2021

Linear Chain Graphical Model


It is well known in the literature that linear chain graphical model can be solved in Polynomial time

a1 a2 a3 a4

This expression needs Non-Deterministic


𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎 = ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )
Polynomial time to evaluate
, , ,

Can be factorized into

This expression needs Polynomial time to


𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎 = ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 )
evaluate

Tree Structure Graphical Model


a1
It is well known in the literature that tree
structure graphical model can be solved in
Polynomial time
a2 a3

a4 a5 a6 a7

This expression
𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎 ,𝑎 ,𝑎 ,𝑎 = ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 ) needs Non-
, , , , , , Deterministic
Polynomial time
Can be factorized into to evaluate
𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎 ,𝑎 ,𝑎 ,𝑎
This expression
= ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 ) needs Polynomial
time to evaluate
26/6/2021

Transform any Boolean algebra so that only


the first layer has ‘Not’ operations
Propagate the ‘Not’ operations to the previous
a1 layer using DE Morgan’s theorem

a2

a3
𝐻
Output
a4

And operation
a5

a6 Or operation
Not operation
at input or
output

Propagate the ‘Not’ operations to the previous


a1 layer using DE Morgan’s theorem

a2

a3
𝐻
Output
a4

And operation
a5

a6 Or operation

Not operation
at input or
output
26/6/2021

Propagate the ‘Not’ operations to the previous


a1 layer using DE Morgan’s theorem

a2

a3
𝐻
Output
a4

And operation
a5

a6 Or operation
Not operation
at input or
output

Propagate the ‘Not’ operations to the previous


a1 layer using DE Morgan’s theorem

a2

a3
𝐻
Output
a4

And operation
a5

a6 Or operation

Not operation
at input or
output
26/6/2021

Propagate the ‘Not’ operations to the previous


a1 layer using DE Morgan’s theorem

a2

a3
𝐻
Output
a4

And operation
a5

a6 Or operation
Not operation
at input or
output

Until only the first layer has ‘Not’ operations


a1

a2 Note that the number


of gates is at most 2
a3 times the original
𝐻 Boolean algebra after
Output propagation.
a4 Polynomial time
Boolean algebra
And operation remains Polynomial
a5
time after
Or operation propagation.
a6
Not operation
For details, please read my paper at input or
output
26/6/2021

NP problem cannot be simplified into P


An NP (Non-Deterministic Polynomial) problem cannot be simplified into Polynomial time by a1 a2
factorization

𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎 = ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 ) a4


a3
, , ,

One possible factorization is shown below. There are many possible different factorizations

𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎 = ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 ) ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )


,

Because of these 3 terms, it takes NP time to evaluate

Why NP problem cannot be simplified

a1 a2 a1 a2 a1 a2

a3 a4 a3 a4 a3 a4

If we remove constraints If we remove constraints However if no constraint


such that the Graphical such that the Graphical is removed, it can never
model becomes a linear model becomes a tree, it be solved efficiently in
chain, it can be solved can be solved efficiently Polynomial time
efficiently in Polynomial in Polynomial time
time
26/6/2021

Boolean algebra representation of NP


problem
The NP Markov Random Field can be expressed as a Boolean algebra
𝑓 𝑎 ,𝑎 ,𝑎 ,𝑎
= ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )ℎ(𝑎 , 𝑎 )
, , ,
=∨ , , , (ℎ 𝑎 , 𝑎 ∧ℎ 𝑎 , 𝑎 ∧ℎ 𝑎 , 𝑎 ∧ℎ 𝑎 , 𝑎 ∧ℎ 𝑎 , 𝑎 ∧ℎ 𝑎 , 𝑎 )

From the previous example that every Boolean algebra can be expressed with only first layer contains ‘Not’ operations, and
this NP Markov Random Field Boolean algebra has no ‘Not’ operation, the simplification of this NP Boolean algebra can be
done without any ‘Not’ operation. (For details, please read my paper)

2sat problem
• There is a special case for the fully connected Graphical a1 a2
model
• If each variable ai can only takes 2 values (0 or 1)
• Then it is a 2sat problem
• The Graphical Model can be solved in Polynomial time a3 a4

• It can be solved by dynamic programming


• 𝐻 𝑎 ,𝑎 = 𝐻 𝑎 ,𝑎 Note that for Hl() and Hl-1(),
the superscripts are indexes
• 𝐻 𝑎 , 𝑎 = 𝐻 (𝑎 , 𝑎 ) ∑ 𝐻 𝑎 ,𝑎 𝐻 (𝑎 , 𝑎 ) and do not mean H raised to
• where 1 ≤ 𝑙 ≤ 𝑛 the power of l
• n is the number of variables 𝑎 , 𝑎 , … , 𝑎

• For details, please read my paper


26/6/2021

Share my links
• I am a Small Person with Big Dreams
• Please help me to repost my links to other platforms so that I can spread my ideas to
the rest of the world

• 我人小,但因梦想而伟大。
• 请帮我的文件链接传发到其他平台,让我的思想能传遍天下。

• Comments? Send to singkuangtan@gmail.com

• Link to my paper
• Prove Np not equal P using Markov Random Field and Boolean Algebra
Simplification
• https://vixra.org/abs/2105.0181

You might also like