An overview of MOSEK Erling D. Andersen MOSEK ApS Fruebjergvej 3, - - PowerPoint PPT Presentation

http://www.mosek.com

An overview of MOSEK

Erling D. Andersen MOSEK ApS Fruebjergvej 3, Symbion Science park, Box 16 2100 Copenhagen O Denmark Email: sales@mosek.com WWW: http://www.mosek.com

October 5, 2013

Introduction

2 / 25

Overview

Introduction Overview Background Product overview Customers Product details Getting started Licensing Present and future

• f MOSEK

Summary

3 / 25

1. Background. 2. Product overview. 3. Customers. 4. Product details. 5. Licensing. 6. Getting started. 7. The future. 8. Summary.

Background

4 / 25

The company

Introduction Background The company Product overview Customers Product details Getting started Licensing Present and future

• f MOSEK

Summary

5 / 25

■

MOSEK ApS is a Danish company.

■

Vision: Create and sell software for solving linear and convex optimization problems.

■

Located in Copenhagen at Symbion Science Park.

■

Daily management: Erling D. Andersen.

■

Currently 4 full-time, 4 part-time employees and 1 phd student.

Product overview

6 / 25

MOSEK optimization tools

Introduction Background Product overview MOSEK

• ptimization tools

Customers Product details Getting started Licensing Present and future

• f MOSEK

Summary

7 / 25

■

Main product: MOSEK optimization tools.

■

Solves generic:

◆

Linear, quadratic, and nonlinear problems. Only convex cases.

◆

Conic optimization problems. min cT x st. Ax = b, x ∈ K where K is a convex cone (only linear, quadratic, semidefinite).

◆

Integer optimization problems.

■

Same as above but some variables are integer constrained.

■

The software is NOT application specific.

Customers

8 / 25

Types of customers

Introduction Background Product overview Customers Types of customers Product details Getting started Licensing Present and future

• f MOSEK

Summary

9 / 25

■

Structural design.

◆

Arch Bridge Analysis.

■

Financial industry.

◆

Portfolio optimization.

◆

Trading and cash optimization.

■

Power industry.

■

Forestry.

■

Bus crew scheduling.

■

Analogue circuit design.

■

Bio tech.

■

Advertisement.

■

+ more.

Introduction Background Product overview Customers Types of customers Product details Getting started Licensing Present and future

• f MOSEK

Summary

10 / 25

■

End customer size.

◆

Small with 1 user.

◆

Large with users e.g. large banks.

■

OEM.

◆

Integrate MOSEK into a another product.

■

Value added resellers. Geographical markets:

■

North America.

■

Europa (few Danish customers).

■

Japan.

■

Australia.

■

South America.

Product details

11 / 25

Optimizers

Introduction Background Product overview Customers Product details Optimizers Interfaces Getting started Licensing Present and future

• f MOSEK

Summary

12 / 25

■

Primal and dual simplex for linear problems.

■

Primal network simplex for network flow problems.

■

Interior-point for linear, conic and nonlinear problems.

■

Two branch and bound and cut optimizers.

◆

Freely included. Tuned for conic problems.

◆

Paid addon: Tuned for linear problems.

Interfaces

Introduction Background Product overview Customers Product details Optimizers Interfaces Getting started Licensing Present and future

• f MOSEK

Summary

13 / 25

■

Optimizer API

◆

Matrix orientated.

◆

C, Java, .NET and Python.

■

Fusion

◆

Object orientated.

◆

Only linear and conic problems.

◆

Java, MATLAB, .NET and Python.

■

Other

◆

AMPL

◆

MATLAB toolbox.

◆

R package.

■

Third party links:

◆

AIMMS, CVX, GAMS, MPL (soon), Woodstock.

◆

Coin OSI, Raven Toolbox, Yalmip, ...

Getting started

14 / 25

Useful links

Introduction Background Product overview Customers Product details Getting started Useful links Licensing Present and future

• f MOSEK

Summary

15 / 25

■

Software download:

◆

http://mosek.com/download/

■

Trial license:

◆

http://mosek.com/resources/trial/

■

Documentation:

◆

http://docs.mosek.com

■

Support:

◆

http://mosek.com/support/

■

Social medias:

◆

http://mosek.com/resources/social-media/

Licensing

16 / 25

Floating licenses

Introduction Background Product overview Customers Product details Getting started Licensing Floating licenses Server licenses Group license Academic license Present and future

• f MOSEK

Summary

17 / 25

■

PTS: Linear and quadratic optimization + integer variables.

■

PTON: Nonlinear and conic extension + integer variables.

■

PTOM: Mixed-integer optimizer extension.

■

One usage at any computer in the network. Token server required.

■

Pricing:

◆

http://mosek.com/sales/pricing/

Server licenses

Introduction Background Product overview Customers Product details Getting started Licensing Floating licenses Server licenses Group license Academic license Present and future

• f MOSEK

Summary

18 / 25

■

PTS-NODE: Linear and quadratic optimization + integer variables.

■

PTON-NODE: Nonlinear and conic extension + integer variables.

■

PTOM-NODE: Mixed-integer optimizer + integer variables.

■

No token server.

■

Unlimited use on a single prespecified computer.

■

Pricing:

◆

http://mosek.com/sales/pricing/

Group license

Introduction Background Product overview Customers Product details Getting started Licensing Floating licenses Server licenses Group license Academic license Present and future

• f MOSEK

Summary

19 / 25

■

A rental license.

■

Unlimited use of 1 or more features.

■

One organizational unit e.g. one department.

■

No license token server.

■

Price negotiated every third year.

Academic license

Introduction Background Product overview Customers Product details Getting started Licensing Floating licenses Server licenses Group license Academic license Present and future

• f MOSEK

Summary

20 / 25

■

Free for both personal and classroom usage.

■

Visit:

◆ http://mosek.com/resources/academic-license/

Present and future of MOSEK

21 / 25

Present

Introduction Background Product overview Customers Product details Getting started Licensing Present and future

• f MOSEK

Present The future Summary

22 / 25

■

MOSEK version 7 released sping 2013.

◆

Added support semidefinite optimization.

◆

Added the linear and conic Fusion modeling API.

◆

Added a new mixed integer optimizer.

■

Strong emphasize on linear and conic problems.

The future

Introduction Background Product overview Customers Product details Getting started Licensing Present and future

• f MOSEK

Present The future Summary

23 / 25

■

Continued strong emphasize on linear and conic

• ptimization + integer variables.

■

Emphasize improvements:

◆

Improve speed and stability.

◆

Exploit hardware better. E.g. AVX and Intel

■

New features:

◆

Add mixed-integer semidefinite optimization.

◆

Add nonsymmetric cones (if algorithms deemed good enough).

Summary

24 / 25

Final comments

Introduction Background Product overview Customers Product details Getting started Licensing Present and future

• f MOSEK

Summary Final comments

25 / 25

■

Presented MOSEK.

■

Slides are available at:

◆

http://www.mosek.com/presentations/