Due to recent updates, all users are required to create an Altair One account to login to the RapidMiner community. Click the Register button to create your account using the same email that you have previously used to login to the RapidMiner community. This will ensure that any previously created content will be synced to your Altair One account. Once you login, you will be asked to provide a username that identifies you to other Community users. Email us at Community with questions.
Constraint/tell neural network only certain outputs possible?
Contributor I
I am trying to build a supervised learning model that predicts the best move for a simplified game similar to tic-tac-toe (i.e. 3 in a row). I created a sample data set where I supplied the best move in form of the integer to play.
Of course, only some outputs (unused spaces) are valid outputs. Is it possible to "tell the NN" that only some outputs are valid in a given situation or constrain the output in some way? Perhaps changing from regression to classification problem but it still doesn't solve that only some outputs should be valid for any given input (i.e. the free squares).
I guess if one were to use this implementation in an "actual game" you could do a post-processing step where if the predicted move is invalid to default to a valid guess. Still, it feels like being able to provide constraints should be natural.
I do not think this is possible, in any regards.
Of course, only some outputs (unused spaces) are valid outputs. Is it possible to "tell the NN" that only some outputs are valid in a given situation or constrain the output in some way? Perhaps changing from regression to classification problem but it still doesn't solve that only some outputs should be valid for any given input (i.e. the free squares).
I guess if one were to use this implementation in an "actual game" you could do a post-processing step where if the predicted move is invalid to default to a valid guess. Still, it feels like being able to provide constraints should be natural.
I do not think this is possible, in any regards.
0
