Online Services

The system provides all types of optimization services:

1. Solving general constrained and unconstrained continuous nonlinear optimization problems.
2. Solving mixed integer nonlinear optimization problems.
3. Solving black-box nonlinear optimization problems.
4. Constructing high-quality neural network ensembles for pattern classification and prediction.

Current system supports solving problems of sizes (expandable):

1. For neural network learning tasks: product of input and output dimensions up to 5,000, sample size up to 2,000.
2. For optimization problems: the number of variables up to 10,000, the number of Jacobian nonzero elements up to 200,000, Hessian nonzero elements up to 1,000,000.

A comprehensive set of interface methods is available for convenient connections to GOT services:

1. Web user interface for task and account management;
2. Standalone client program for more responsive interaction;
3. Service API for C/C++, JAVA, MATLAB, Python programs;
4. Connection to cloud compute services (e.g. Amazon E2C) for unlimited, on-demand system scalability.

Comprehensive model support:

1. Supports problem types: unconstrained and constrained, linear and nonlinear, continuous, mixed-integer and combinatorial optimization problems.
2. Supports a comprehensive set of solvers (MINPACK, IPOPT, all GAMS solvers) with automatic and intelligent solver selection for the submission.
3. Supports problem submissions formatted in GOT, GAMS, AMPL, ZIMPL model specifications.