June 04, 2004
Model Details Paper Released
|
After a meeting yesterday with Henry, I think that we convinced ourselves that we have the details of the activity inference model worked out. For the benefit of the hordes of people who would like to know the structure of the DBN that we are proposing, how an activity graph is compiled to a DBN and the conditional probability tables of that DBN, I present the following file. |
Here is an interesting question: Are DBNs a subset of HMMs or a superset?
I call DBNs a subset of HMMs because DBNs represent a smaller or equal set of sub-state dependencies than an HMM (and gain an increase in efficiency as a result). In the worst case an HMM would have to have an exponentially larger number of conditional probability table entries than a DBN to represent the same dependencies. At the end of the day though, an HMM can express relationships that if a DBN were to express the same, would require the DBN to be as large as the HMM and so I say a DBN is a subset.
Another good question: Where are the partial orders?
The answer, the technical write-up is lagging behind the implementation right now. This is the math from the linear temporal ordering stage. The partial orders will come eventually.