One conceptual difference is that BN corresponds to a "third person" viewpoint on probable relations among events, while MDP corresponds to a "first person" viewpoint on probable relations between actions and consequences.
Pei On 6/24/07, Bo Morgan <[EMAIL PROTECTED]> wrote:
My understanding is that MDPs and Bayes Nets are basically slightly different ways to represent probabilistic transitions in finite state machines. Again in my understanding, which please correct if I'm incorrect, is that the primary advantage of Bayes Nets is that you can remove a portion of the graph and still have it be a well defined probability density, whereas this is not possible with the Markov Random Field that implements the Markov Decision Process. However, this doesn't sound like much of a difference in terms of big organizational ideas. This sounds like differences in implementation details. Bo On Sun, 24 Jun 2007, Lukasz Stafiniak wrote: ) Ouch, they differ more than I thought... Good :-) ) ) (HTM based more on Bayes nets) ) ) On 6/24/07, Lukasz Stafiniak <[EMAIL PROTECTED]> wrote: ) > I'm starting to learn about Numenta's HTM, but perhaps someone would ) > like to share in advance: ) > what are the essential differences between HTM and Yuang Weng's IHDR ) > augmented with Observation-driven MDPs? ) > ) ) ----- ) This list is sponsored by AGIRI: http://www.agiri.org/email ) To unsubscribe or change your options, please go to: ) http://v2.listbox.com/member/?&; ) ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?&;
----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=231415&user_secret=e9e40a7e
