Optimization shell Inverse has undergone the first development stage where the basic framework for implementation tools have been developed. It is now in the stage where algorithms and tools for performing different tasks and solving various kinds of sub problems are added and improved. New tools will be mostly responsible for changing the programme look, although some conceptual things will be the subject of change and upgrade, too.

    There are many things we would like to do in the future. Inverse tends to be a programme offering tools for various optimization areas such as shape optimization, optimization of systems with noisy response, etc.  We would not like to limit Inverse to the use with a single simulation programme or to solution of a specific type of problems. One of the basic ideas of the shell is generality. It should in the first place provide general tools commonly needed for solution of optimization problems, e.g. optimization algorithms or functions for matrix manipulation. On the top of this it should provide a framework for implementation of tools for solution of specific sub problems (e.g. domain transformation for shape optimization problems and interfacing routines for utilization of simulation programmes).

    One of the objectives for the future is to provide Inverse as an open library.  Different research groups will then be able to implement their own tools in Inverse independently on the Inverse development team. Modules developed in different research and development groups can this way be distributed without interference of Inverse development team. We hope that gradually this would create variety of tools that users could obtain from different sources and plug into Inverse. The Inverse team will however still develop its own tools which will be a part of the shell by default. We want to provide a set of most basic tools to cover common needs, but also leave peak experts in individual areas to offer their more advanced and specialized solutions as a supplement to the basic scheme.

    In a similar way we would like to spread a circle of simulation codes that would provide a close interface with Inverse. All users of these codes would have, regardless to the simulation code they use, opportunity to take the advantage of what Inverse offers, essentially of a variety of tools developed not only by the Inverse team. Inverse comes with a general file interface, which enables use of Inverse with practically every code that uses text input and output. However, direct interface between the shell and a simulation programme is more elegant, efficient and easy to use. Besides, the structure of many simulation programme is not suited wery well for optimization. Creation of a direct interface with Inverse can improve that. Inverse team will provide assistance to the simulation code developers who would like to provide a direct interface with Inverse. The expertise of the whole C3M group, a part of which is doing research in advanced finite element code concepts, can be employed here.

    Development tasks that currently have the highest priority are the following:

• Optimization algorithms - the FSQP represents the strongest optimization engine of the shell. It serves wery well for identification of local minima when the objective function is enough smooth and the level of noise is not too high. In the future the shell will have to provides efficient algorithms for problems for which standard algorithms don't perform well, e.g. statistical optimization methods. Implementation of such algorithms can generate additional demands on shell concepts which Inverse development team will have to meet. The beginning stage of incorporating new kinds of algorithms will be crucial for identifying such demands.
• Open library - this will enable independent development of additional modules outside the Inverse development team. Basic task will be development of library functions which will support complete integration of newly developed modules with the existent shell functionality.
• Parallelisation - the parallel interface will provide the framework for implementation of  parallel optimization algorithms.
• Facilities for processing results, including graphics - standard data formats will be provided for representation of optimization results, which include not only the optimal parameter set and function values, but also optimization path. These structures will be generated by optimization algorithms, while the user will be able to customize the structures and output the data in various formats which will enable additional processing by external programmes. A graphic module will provide graphical interpretation of data in the standard format.
• Auxiliary tools, e.g. matrix manipulation or supporting tools for shape optimization, will be continuously developed.
• Integration with simulation programmes will not only make use of Inverse with specific simulation programmes easier, but will also accumulate knowledge about demands on the structure of simulation software from different areas that arise from optimization.

    Inverse development has depended much on funding granted for research. This source will not suffice in the future, therefore we will send the programme to market. With the additional source of income generated by commercial distribution we will try to build a research and development team which will be up to meeting the most demanding user needs.