Hello everybody,
I am looking for a learning paradigm (and the system / framework / network that goes with it), in order to accompish the following: Consider an agent evolving in a 2D/3D world, and at each time step we have a compact representation of its input in the form of 1000s binary variables (*X1...X1000.*), with each binary variable potentially representing a piece of knowledge about the input. Now, imagine that there is a temporal causal relation between *X1* and *X9*, i.e when *X1* is 1 at time step *t*, then *X9* becomes 1 at time step *t+1* *So my question is*: how to capture this kind of propositional knowledge ? I have been thinking about various ML paradigms such as Markov Logic Network (each node is of the FO type), or Dynamic Bayesian Network (allows to represent temporal dependencies btween nodes), But I was wondering if a Probabilistic Logic Network was not more suited. One of the problem that the framework should solve in particular is that: *X1 -> X9 *(between two consecutive time steps) can be seen very *sparsly *(because for ex: X1 = true will be rarely true while the agent evolves), and yet we "know" that X1 -> X9. Best Aymeric -- You received this message because you are subscribed to the Google Groups "opencog" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/b83867a3-c389-4c3f-96dc-065fff6501d6n%40googlegroups.com.
