Version 1.1.8

• Simplified barrier parameters in augl.
• Added problem properties and methods for checking whether solvers support certain problem properties.
• Added methods to convert general problem into linear, mixed integer linear, and quadratic problems.
• Added methods to write lp files from linear and mixed integer linear problems.
• Added command-line interfaces for cbc and clp.
• Added command-line interface for cplex.
• Added acceleration factor to nr.
• Added recovery strategy for augl linear system factorization error.
• Lagrange multiplier extraction from cbc/clp/cplex solutions.

Version 1.1.7

• Got CBC interface working again on Linux.
• Improved location of inlp termination condition in order not to miss output of last iter.
• Added version file.

Version 1.1.6

• Fixed bug with Linux installation with OPTALG_IPOPT=true.
• Made AugL/INLP/Ipopt work with empty problems.
• Added line_search option to INLP as well as line search maxiter param.
• Added UMFPACK linear solver interface (can be used if scikit-umfpack is installed).
• Changed 'linsolver' param value of nr to 'default'.
• INLP improvements: separate primal and dual steps, better var initializations.
• Fixed IPOPT windows dll handling.
• Seperated feasibility and optimality tolerances in inlp.

Version 1.1.4

• Improved error checks in ipopt wrapper, and added derivative_test and hessian approx option.
• Added linear_solver and print_level options for ipopt.
• IQP allows general problem as input, forms QP approximation.
• Naive, but hopefully efficient interior-point nonlinear programming solver (inlp) based on taking newton steps on the KKT conditions.
• Improved obj scaling of IQP (to match that of INLP).
• Improved bound stretching in AugL for problems with non-empty interior.

Version 1.1.3

• Portable setup.py (--with argument).
• Linear problem class.
• Mixed integer linear problem class.
• Coin-OR Clp interface.
• Coin-OR Cbc interface.
• Auto objective scaling in AugL solver.
• IPOPT wrapper extracts number of iterations.