Skip to main content

Table 1 Complexity results for the eligible kernel functions

From: Kernel-function-based primal-dual interior-point methods for convex quadratic optimization over symmetric cone

i

The eligible kernel functions ψ i (t)

Large-update methods

Small-update methods

Ref.

1

t 2 − 1 2 −logt

O(rlog r ε )

O( r log r ε )

e.g., [1]

2

1 2 ( t − 1 t ) 2

O( r 2 3 log r ε )

O( r log r ε )

[6]

3

t 2 − 1 2 + t 1 − q − 1 q − 1 , q>1

O(q r q + 1 2 q log r ε )

O( q 2 r log r ε )

[6]

4

t 2 − 1 2 + t 1 − q − 1 q ( q − 1 ) − q − 1 q (t−1), q>1

O(q r q + 1 2 q log r ε )

O( q 2 r log r ε )

[5]

5

t 2 − 1 2 + e 1 t − e e

O( r ( log r ) 2 log r ε )

O( r log r ε )

[6]

6

t 2 − 1 2 − ∫ 1 t e 1 ξ − 1 dξ

O( r ( log r ) 2 log r ε )

O( r log r ε )

[6]

7

t 2 − 1 2 + e q ( 1 t − 1 ) − q q , q ≥ 1

O(q r log r ε )

O(q q r log r ε )

[7]

8

t 2 − 1 2 − ∫ 1 t e q ( 1 ξ − 1 ) dξ, q ≥ 1

O(q r log r ε )

O(q q r log r ε )

[6]

9

t 2 − 1 2 + ( e − 1 ) 2 e 1 e t − 1 − e − 1 e

O( r 3 4 log r ε )

O( r log r ε )

[10]

10

8 t 2 −11t+1+ 2 t −4logt

O( r 5 6 log r ε )

O( r log r ε )

[19]

11

8 t 2 −10t+ 2 t 3

O( r 5 8 log r ε )

O( r log r ε )

[14]

12

t 2 − 1 2 + 6 π tan( π ( 1 − t ) 2 + 4 t )

O( r 3 4 log r ε )

O( r log r ε )

[17]

13

t 2 − 1 2 −log(t)+ 1 8 tan 2 ( π ( 1 − t ) 2 + 4 t )

O( r 2 3 log r ε )

O( r log r ε )

[21]

14

p ( t 2 − 1 ) 2 + t − p q − 1 q ( q + 1 ) − p q ( t − 1 ) q + 1 , p ≥ 1, q>0

O( r logrlog r ε )

O( r log r ε )

[15]

15

t+ 1 t −2

O(rlog r ε )

O( r log r ε )

[9]

16

t−1+ t 1 − q − 1 q − 1 , q>1

O(qrlog r ε )

O( q 2 r log r ε )

[6]

17

t p + 1 − 1 p + 1 −logt, p∈[0,1]

O(rlog r ε )

O( r log r ε )

[18]

18

{ t p + 1 − 1 p + 1 + t 1 − q − 1 q − 1 , t > 0 , p ∈ [ 0 , 1 ] , q > 1 t p + 1 − 1 p + 1 − log t , t > 0 , p ∈ [ 0 , 1 ] , q = 1

O(q r p + q q ( 1 + p ) log r ε )

O( q 2 r log r ε )

[8]