Oh man, Florian Fuchs is really going to kill me for this, but this blog post had completely gone off my mind. Anyway, so here I am writing this post long after it was supposed to be written. So I will just write about the experiences which I had till 13th. (They were interesting experiences for sure). I will also pretend in the blog as if it is 13th today :p
So, starting from where I had finished off in the last post. I was left with a broken factor function and some frustration arising from that. I had posted the issue on our group, but Shashank and Ankur were busy with other stuff and they were not finding time to solve it. So in the last GSOC chat, Shashank told me to write a new factor function itself. He thought that it would be a good exercise in cython to do that on my own and also, it would be very difficult for me to parse the code which Shashank had initially written for the factor product function.
Also, I decided to include a new list to store the MAP values along with the numpy.array which we had for storing the potential values. (Corresponding to each value, there would be the values which the other variables had taken during elimination). There were a lot of tough design decisions here (how to store it : list vs dictionary. Is a list fine? However, I implemented each of the entries as a list of the eliminated variable and stored the entire thing in a list. I am not really sure what to do about the huge expense of the fact that we are storing it in a list. For later.
After this, I pinged shashank again about the product module since it was making testing very difficult for me. So shashank told me to implement it myself since it would be a very good exercise in learning cython for me too. So I implemented this in cython and it was working fine. However, there was this big issue. How to handle the eliminated variable values in the cython function. because they were stored in lists and we could only pass C++ data structures to the cython functions. I am on the lookout for possible ways to handle this.
Anyway, that's all for now. (from the point of view of 13th August).
So, starting from where I had finished off in the last post. I was left with a broken factor function and some frustration arising from that. I had posted the issue on our group, but Shashank and Ankur were busy with other stuff and they were not finding time to solve it. So in the last GSOC chat, Shashank told me to write a new factor function itself. He thought that it would be a good exercise in cython to do that on my own and also, it would be very difficult for me to parse the code which Shashank had initially written for the factor product function.
Also, I decided to include a new list to store the MAP values along with the numpy.array which we had for storing the potential values. (Corresponding to each value, there would be the values which the other variables had taken during elimination). There were a lot of tough design decisions here (how to store it : list vs dictionary. Is a list fine? However, I implemented each of the entries as a list of the eliminated variable and stored the entire thing in a list. I am not really sure what to do about the huge expense of the fact that we are storing it in a list. For later.
After this, I pinged shashank again about the product module since it was making testing very difficult for me. So shashank told me to implement it myself since it would be a very good exercise in learning cython for me too. So I implemented this in cython and it was working fine. However, there was this big issue. How to handle the eliminated variable values in the cython function. because they were stored in lists and we could only pass C++ data structures to the cython functions. I am on the lookout for possible ways to handle this.
Anyway, that's all for now. (from the point of view of 13th August).
No comments:
Post a Comment