- 25 Feb 2019, 11:33
- Forum: General Discussion
- Topic: Enumerating all integer feasible points, CPLEX vs Polymake
- Replies:
**18** - Views:
**1216**

./configure --prefix=/opt/ppl --enable-ppl_lcdd This would usually require sudo for make install, because /opt is owned by root. Thanks for pointing this out. I typiclaly chown the folders before, e.g.: sudo mkdir /opt/ppl sudo chown opfer. /opt/ppl Then you don't need sudo for the installation pro...

- 25 Feb 2019, 10:37
- Forum: General Discussion
- Topic: Enumerating all integer feasible points, CPLEX vs Polymake
- Replies:
**18** - Views:
**1216**

First, please install M4: apt install m4 Then see if configure has further errors. Also, on Ubuntu, I would not install to the default path. I'd rather do something like ./configure --prefix=/opt/ppl --enable-ppl_lcdd Then after "make" and "make install", you should be able to call /opt/ppl/bin/ppl_...

- 25 Feb 2019, 10:19
- Forum: General Discussion
- Topic: Enumerating all integer feasible points, CPLEX vs Polymake
- Replies:
**18** - Views:
**1216**

Thomas, Thanks. But it is not working for me :-( I downloaded ppl from their website, then did configure with the enabling on, make and install. Did you use the parameter "--enable-ppl_lcdd" for configure? polymake does not use ppl in the floating-point mode yet, rather for rational numbers only. T...

- 25 Feb 2019, 09:13
- Forum: General Discussion
- Topic: Enumerating all integer feasible points, CPLEX vs Polymake
- Replies:
**18** - Views:
**1216**

Thomas, I am simply unable to get such results at my end. Could you please share with me the steps you take to get the facetial representation in 20 seconds? Yes, sure. But as I said, this was just a test (completely without polymake). Here is what I did: 1. Configure, compile and install PPL (conf...

- 24 Feb 2019, 21:21
- Forum: General Discussion
- Topic: Enumerating all integer feasible points, CPLEX vs Polymake
- Replies:
**18** - Views:
**1216**

I just did another test: I used PPL (Parma Polyhedra Library) to calculate the H-representation from these 4722 points. It took around 20 seconds to generate 414 exact rational (in)equalities. Do you know whether this number is correct?

Best regards,

Thomas

Best regards,

Thomas

- 24 Feb 2019, 18:53
- Forum: General Discussion
- Topic: Enumerating all integer feasible points, CPLEX vs Polymake
- Replies:
**18** - Views:
**1216**

Is there any software you are aware of that computes facetial inequality and equations of the affine hull using faster but less accurate floating point arithmetic when it is provided with the set of lattice points? As far as I know, both cdd and lrs should be able to do this. But you should also be...

- 23 Feb 2019, 18:57
- Forum: General Discussion
- Topic: Enumerating all integer feasible points, CPLEX vs Polymake
- Replies:
**18** - Views:
**1216**

I think this has 2 reasons: 1. polymake uses exact arithmetic. Exact calculations are much slower than floating point arithmetic used in CPLEX. 2. If I remember correctly, polymake currently uses convex hull algorithms to enumerate the lattice points. A year ago, I wrote a Cut&Branch algorithm that ...

- 19 Apr 2018, 20:46
- Forum: General Discussion
- Topic: facets of a mixed ILP
- Replies:
**7** - Views:
**3008**

Yes, perhaps I should look at projecting out the continuous variables. Some of these variables are /implicitly/ binary though. Will there a performance gain in polymake if I explicitly declare in the .lp file variables to be binary (that are implicitly so)? I cannot answer this for sure, but I thin...

- 19 Apr 2018, 07:39
- Forum: General Discussion
- Topic: facets of a mixed ILP
- Replies:
**7** - Views:
**3008**

Well, I have to admit that I don't know whether there is an algorithm for calculating mixed-integer polytopes in polymake. We will have to ask my colleagues. In the tutorial that you mentioned, the continuous variables are projected out. Do you want to keep them or do you want to project them out? B...

- 18 Apr 2018, 20:57
- Forum: General Discussion
- Topic: facets of a mixed ILP
- Replies:
**7** - Views:
**3008**

Hi, thanks for the LP file. Could you please elaborate a little on which information you need. Solving the LP and iterating the integer solutions seems rather easy. There are only 4 different feasible assignments for the integer variables. Are you looking for facets that belong to the convex hull of...

Powered by phpBB® Forum Software © phpBB Limited