Disclaimer: Dieser Thread wurde aus dem alten Forum importiert. Daher werden eventuell nicht alle Formatierungen richtig angezeigt. Der ursprüngliche Thread beginnt im zweiten Post dieses Threads.
The mysterious fortune cookie
In assignment 7.1.2 applying smoothing was simplified by providing the base case of the BACKWARD algorithm from section 22.2.
My question is: how does one obtain these values for the base case? I tried setting t to k+1 in the derivation of the BACKWARD algorithm (slide 771), yielding P(e_{k+1:k+1} | X_k) = \sum_{x_{k+1}} P(e_{k+1:k+1} | x_{k+1}) * P(x_{k+1} | X_k). Since e_{k+1:k+1} = e_{k+1}, one obtains \sum_{x_{k+1}} P(e_{k+1} | x_{k+1}) * P(x_{k+1} | X_k), which can be computed using the sensor and transition model respectively.
When I tried this computation, I got different values from the fortune cookie however. Can someone help?