1 of 12

On the Hardness of Compressing Weights

Bart M. P. Jansen Eindhoven University of Technology�Shivesh K. Roy Eindhoven University of TechnologyMichał Włodarczyk Eindhoven University of Technology

MFCS 2021�Aug 23-27, 2021, Tallinn (Estonia)

2 of 12

Data reduction with a guarantee for NP-hard problems

  •  

2

 

 

 

 

 

 

 

A kernelization guarantees that instances that are large �with respect to their complexity parameter can be shrunk

3 of 12

Unweighted vs. Weighted problems

  •  

3

4 of 12

Weights in kernelization

  •  

4

Is this overhead in the kernel size of weighted problems necessary?

?

5 of 12

Exact-Edge-Weight Clique (EEWC)

  •  

5

t= 21

6 of 12

Exact-Edge-Weight Clique (EEWC)

  •  

6

7 of 12

Our results for Exact-Weighted problems:

7

Problem

Parameter

Kernel upper bound

(Randomized)

Exact-Edge-Weight Clique

(EEWC)

#vertices n

#vertices n

#vertices n

?

 

#variables n

Subset Sum

#items n

Matching upper bound is known

[Harnik & Naor, SICOMP]

Tightens result of Etscheid et al. [JCSS] who ruled out standard kernelization assuming ETH

8 of 12

Our results for Weighted Vertex Cover

  •  

8

9 of 12

Towards cubic lower bound for EEWC

  •  

9

 

10 of 12

Key lemma

10

 

 

 

 

 

 

 

 

 

Hypothetical Kernel

 

 

 

11 of 12

 

11

 

 

 

# vertices added

 

 

 

 

 

 

 

 

 

 

 

 

 

12 of 12

Conclusion

  •  

12

THANK YOU!