Partnering

R&D is going on to train Combinatorial AI to

• Process structured problem descriptions in extentions of XML
• Represent solutions in a language that is Turing-complete

We look for application-oriented strategic partners interested to train Combinatorial AI in a specific application domain.

There’re three facets of knowledge to be maintained in a development project:

• Languages
• Understanding problems
• Coding skills

A newer problem understanding usually requires new concepts, which leads to development of languages used for formal definitions of the problems. This part of work is the field for humans.

Languages for solutions are often fixed in a project and give little or no room for development. These languages are usually learned “as is”, without questions on how they can be changed. This kind of coding is the right role for the combinatorial agent.

Problems are usually defined using descriptive languages, which can be extentions of XML or other languages adopted for the development needs in your team. (Note that many data file formats of modern applications are XML-based).

There is no difference for the agent between training on model problems and solving real problems. The agent can learn an additional concept or language on model problems first and use this knowledge on the next iteration of develpment.

As development continues, the language for problem definitions can be developed along with your understanding of the problems.

• A training scheme for the combinatorial agent consisting of some model problems from the application domain is implemented in the reward unit;
• The scheme is used to train the agent;
• After the training, the combinatorial agent continues looking for solutions for other user-defined problems from the application domain.

• Choice of an application domain
• Choice of the input language to define problems and the output language for the solutions
• Training the agent to solve the applied problems

• Elaboration of agent’s communication with the training environment and reward units
• Training the agent to use the specified languages