# American Institute of Mathematical Sciences

• Previous Article
Scheduling jobs with controllable processing time, truncated job-dependent learning and deterioration effects
• JIMO Home
• This Issue
• Next Article
Infinite-time ruin probability of a renewal risk model with exponential Levy process investment and dependent claims and inter-arrival times
April  2017, 13(2): 1009-1024. doi: 10.3934/jimo.2016059

## A superlinearly convergent hybrid algorithm for solving nonlinear programming

 1 School of Economics and Management, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China 2 School of Management, Shanghai University, Shanghai 200444, China 3 School of Mathematics and Physics, Huaiyin Institute of Technology, Huaian, Jiangsu 223003, China

Received  April 2015 Revised  June 2016 Published  August 2016

Fund Project: The research is supported by the National Natural Science Foundation of China (No. 11501350)

In this paper, a superlinearly convergent hybrid algorithm is proposed for solving nonlinear programming. First of all, an improved direction is obtained by a convex combination of the solution of an always feasible quadratic programming (QP) subproblem and a mere feasible direction, which is generated by a reduced system of linear equations (SLE). The coefficient matrix of SLE only consists of the constraints and their gradients corresponding to the working set. Then, the second-order correction direction is obtained by solving another reduced SLE to overcome the Maratos effect and obtain the superlinear convergence. In particular, the convergence rate is proved to be locally one-step superlinear under a condition weaker than the strong second-order sufficient condition and without the strict complementarity. Moreover, the line search in our algorithm can effectively combine the initialization processes with the optimization processes, and the line search conditions are weaker than the previous work. Finally, some numerical results are reported.

Citation: Chuanhao Guo, Erfang Shan, Wenli Yan. A superlinearly convergent hybrid algorithm for solving nonlinear programming. Journal of Industrial & Management Optimization, 2017, 13 (2) : 1009-1024. doi: 10.3934/jimo.2016059
##### References:

show all references

##### References:
The performance of ALGO 1 for solving problems in Table 1 under Case Ⅰ and Case Ⅱ, respectively.
The performance of ALGO 1 for solving problems in Table 2 under Case Ⅰ and Case Ⅱ, respectively.
The left figure shows the performance of ALGO 1-Ⅱ and ALGO-WCH, the right figure shows the performance of ALGO 1-Ⅱ and ALGO 2.1.
The corresponding parameters for the test problems (1)
 Prob $n/m$ $x^0$ $\Upsilon(x^0)$ Prob $n/m$ $x^0$ $\Upsilon(x^0)$ 013 2/3 $(-5,-5)^T$ 5.00 098 6/16 $(20,0,\ldots,0)^T$ 19.69 018 2/6 $(1,5)^T$ 20.00 100 7/4 $(3,\ldots,3)^T$ 188.00 023 2/9 $(-5,-5)^T$ 11.00 108 9/14 $\begin{array}{ll}(0.1,5,0.5,\ 1,0,1,\ 0.5,5,0.1)^T\end{array}$ 23.26 030 3/7 $(5,10,15)^T$ 124.00 224 2/8 $(-2,-1)^T$ 5.00 033 3/6 $(1,4,7)^T$ 2.00 225 2/5 $(1,4)^T$ 3.00 038 4/8 $(0,3,1,2)^T$ 3.00 234 2/5 $(1,3)^T$ 9.00 043 4/3 $(-10,2,-8,5)^T$ 236.00 250 3/8 $(-3,-3,-3)^T$ 15.00 044 4/10 $(-20,\ldots,-20)^T$ 20.00 #38; 264 4/3 $(0,0,0,3)^T$ 6.00 066 3/8 $(0,0,100)^T$ 90.00 337 3/3 $(1,1,0)^T$ 1.00 076 4/7 $(1,2,3,4)^T$ 7.00 354 4/5 $(1,-1,1,-1)^T$ 1.00
 Prob $n/m$ $x^0$ $\Upsilon(x^0)$ Prob $n/m$ $x^0$ $\Upsilon(x^0)$ 013 2/3 $(-5,-5)^T$ 5.00 098 6/16 $(20,0,\ldots,0)^T$ 19.69 018 2/6 $(1,5)^T$ 20.00 100 7/4 $(3,\ldots,3)^T$ 188.00 023 2/9 $(-5,-5)^T$ 11.00 108 9/14 $\begin{array}{ll}(0.1,5,0.5,\ 1,0,1,\ 0.5,5,0.1)^T\end{array}$ 23.26 030 3/7 $(5,10,15)^T$ 124.00 224 2/8 $(-2,-1)^T$ 5.00 033 3/6 $(1,4,7)^T$ 2.00 225 2/5 $(1,4)^T$ 3.00 038 4/8 $(0,3,1,2)^T$ 3.00 234 2/5 $(1,3)^T$ 9.00 043 4/3 $(-10,2,-8,5)^T$ 236.00 250 3/8 $(-3,-3,-3)^T$ 15.00 044 4/10 $(-20,\ldots,-20)^T$ 20.00 #38; 264 4/3 $(0,0,0,3)^T$ 6.00 066 3/8 $(0,0,100)^T$ 90.00 337 3/3 $(1,1,0)^T$ 1.00 076 4/7 $(1,2,3,4)^T$ 7.00 354 4/5 $(1,-1,1,-1)^T$ 1.00
The corresponding parameters for the test problems (2)
 Prob $n/m$ $x^0$ $\Upsilon(x^0)$ Prob $n/m$ $x^0$ $\Upsilon(x^0)$ Svanberg-20 20/60 $(5,\ldots,5)^T$ 4.20 S394-40 40/1 $(0.4,\ldots,0.4)^T$ 5.40 $(0.5,\ldots,0.5)^T$ 2.42 $(0.3,\ldots,0.3)^T$ 5.30 Svanberg-30 30/90 $(5,\ldots,5)^T$ 4.20 S394-70 70/1 $(0.4,\ldots,0.4)^T$ 10.20 $(0.5,\ldots,0.5)^T$ 2.50 $(0.3,\ldots,0.3)^T$ 5.30 Svanberg-40 40/120 $(5,\ldots,5)^T$ 4.20 S394-100 100/1 $(0.4,\ldots,0.4)^T$ 15.00 $(0.5,\ldots,0.5)^T$ 2.54 $(0.3,\ldots,0.3)^T$ 8.00 Svanberg-50 50/150 $(5,\ldots,5)^T$ 4.20 S394-150 150/1 $(0.4,\ldots,0.4)^T$ 23.00 $(0.5,\ldots,0.5)^T$ 2.57 $(0.3,\ldots,0.3)^T$ 12.50 Svanberg-200 200/600 $(5,\ldots,5)^T$ 4.20 S394-200 200/1 $(0.4,\ldots,0.4)^T$ 31.00 $(0.5,\ldots,0.5)^T$ 2.64 $(0.3,\ldots,0.3)^T$ 17.00 Svanberg-300 300/900 $(5,\ldots,5)^T$ 4.20 S394-300 300/1 $(0.4,\ldots,0.4)^T$ 31.00 $(0.5,\ldots,0.5)^T$ 2.65 $(0.3,\ldots,0.3)^T$ 17.00
 Prob $n/m$ $x^0$ $\Upsilon(x^0)$ Prob $n/m$ $x^0$ $\Upsilon(x^0)$ Svanberg-20 20/60 $(5,\ldots,5)^T$ 4.20 S394-40 40/1 $(0.4,\ldots,0.4)^T$ 5.40 $(0.5,\ldots,0.5)^T$ 2.42 $(0.3,\ldots,0.3)^T$ 5.30 Svanberg-30 30/90 $(5,\ldots,5)^T$ 4.20 S394-70 70/1 $(0.4,\ldots,0.4)^T$ 10.20 $(0.5,\ldots,0.5)^T$ 2.50 $(0.3,\ldots,0.3)^T$ 5.30 Svanberg-40 40/120 $(5,\ldots,5)^T$ 4.20 S394-100 100/1 $(0.4,\ldots,0.4)^T$ 15.00 $(0.5,\ldots,0.5)^T$ 2.54 $(0.3,\ldots,0.3)^T$ 8.00 Svanberg-50 50/150 $(5,\ldots,5)^T$ 4.20 S394-150 150/1 $(0.4,\ldots,0.4)^T$ 23.00 $(0.5,\ldots,0.5)^T$ 2.57 $(0.3,\ldots,0.3)^T$ 12.50 Svanberg-200 200/600 $(5,\ldots,5)^T$ 4.20 S394-200 200/1 $(0.4,\ldots,0.4)^T$ 31.00 $(0.5,\ldots,0.5)^T$ 2.64 $(0.3,\ldots,0.3)^T$ 17.00 Svanberg-300 300/900 $(5,\ldots,5)^T$ 4.20 S394-300 300/1 $(0.4,\ldots,0.4)^T$ 31.00 $(0.5,\ldots,0.5)^T$ 2.65 $(0.3,\ldots,0.3)^T$ 17.00