SDPA (Semidefinite Programming Algorithm)

Katsuki Fujisawa, Masakazu Kojima and Kazuhide Nakata

Dear Colleagues: 

We would like to announce a new version of 

``SDPA (Semidefinite Programming Algorithm)''
by Katsuki Fujisawa, Masakazu Kojima and Kazuhide Nakata, 

which can be obtained at

Best wishes, Katsuki Fujisawa
             Masakazu Kojima
             Kazuhide Nakata

The SDPA (SemiDefinite Programming Algorithm) solves the standard
form semidefinite program and its dual. It is coded in C++ utilizing 
the Meschach for matrix computation. The SDPA version 1 was written in 
December 1995 based on the interior-point algorithm given by Kojima, 
Shindoh and Hara for monotone semidefinite linear complementarity 
problems. A considerable revision (see Section 6 of this manual) was 
made in August 1996. The SDPA version 2 enjoys the following features: 

1. Some information on infeasibility of a semidefinite program to 
   be solved is provided. 
2. Block diagonal matrix structure and sparse matrix structure 
   in data matrices are available. 
3. Mehrotra-type predictor-corrector is incorporated.  
4. Three types of search directions are available.

This manual and the SDPA can be found at 

Read the file ``README'' there for more details on how to get the SDPA.

We are grateful to Michael Overton and Mike Todd. They kindly sent us 
their MATLAB codes of their primal-dual interior-point methods for SDPs. 
>From their codes, we learned a lot on how to implement the Mehrotra 
type predictor-corrector step, the Alizadeh-Haeberly-Overton search 
direction and the Nesterov-Todd search direction.