## Homework4

Moderator: Lernende Roboter

doro
Neuling
Beiträge: 2
Registriert: 8. Feb 2014 11:11

### Homework4

We have some problems understanding the task of homework4.
First it is not clear how to do the 1a). We assume we have to derive with respect to w? It would be helpfull if someone could explain how to derive a log of a gaussian inside a sum.
Second we have huge roblems understanding the code framework for 1b).
It does not reveal what the dimensions of theta are and in particular it is unclear how to aply this theta on the robot. What is theta anyway? The actions for the robot?
It would help a lot to explain the structure of the given framework shortly instead of expecting us to do reverse engineering.
And we do not understand if theta_i is the i-th component of theta or not?

doro
Neuling
Beiträge: 2
Registriert: 8. Feb 2014 11:11

### Re: Homework4

and we do not know how to sample so that we get a vector for theta
when mue is a vector of (1,10) and sigma a matrix of 10x10
theta=normrnd(mue,sigma); gives an dimension error.
the only way it is working is this:
theta=diag(normrnd(diag(mue),sigma));
but i this correct ? Or is there aother way of sampling ?

mf1008
Erstie
Beiträge: 15
Registriert: 19. Okt 2008 18:05

### Re: Homework4

doro hat geschrieben:and we do not know how to sample so that we get a vector for theta
when mue is a vector of (1,10) and sigma a matrix of 10x10
theta=normrnd(mue,sigma); gives an dimension error.
the only way it is working is this:
theta=diag(normrnd(diag(mue),sigma));
but i this correct ? Or is there aother way of sampling ?
using mvnrnd() (for generating multivariate normal random numbers) should solve this
And we do not understand if theta_i is the i-th component of theta or not?
theta^ is the i th sample in each iteration