3
$\begingroup$

For example I would like to implement transparent AI in the RTS game which doesn't offer any AI API (like old games), and I'd like to use image recognition algorithm for detecting the objects which can talks to another algorithm which is responsible for the logic.

Given I'd like to use two neural networks, what are the approaches to setup the communication between them? Is it just by exporting result findings of the first algorithm (e.g. using CNN) with list of features which were found on the screen, then use it as input for another network? Or it's more complex than that, or I need to have more than two networks?

$\endgroup$
4
$\begingroup$

The underlying abstraction (which is essentially what you'd be using the first network for) is that of reducing the state-space of the raw input via feature extraction/synthesis and/or dimensionality reduction.

At present, there are few definite rules for doing this: practice is more a question of 'informed trial and error'.

If you add some information to your question regarding what has been previously attempted in this area (e.g. on the ALE platform), this it might be possible to offer some more specific advice.

$\endgroup$
0
$\begingroup$

That depends on what type of network you want to use for your second network, instead of feeding the outputs of the first layer, it would be much better if you jointly train both the networks. But that depends on the architecture of the second network ('logic' network).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.