Weekly Journal: Week 8 (July 9 to July 15)

Victoria Manfredi
vmanfred@cs.smith.edu


I have spent this week debugging the q-learning algorithm. I discovered a lot of bugs, but I also learned a lot about what the qlearning algorithm was doing. I have debugged most of it, but I am tracing through now, collecting data and making graphs trying to make sure that it is correct. One of the bugs was that the buyer proportions of min buyers and any buyers were different than the ones for the nash equilibrium I was looking for in the prices chosen. There were lots of little bugs too. If you are interested in seeing some data that is probably correct :-) see below.

I also spent some time playing with pstogif and ghostscript, trying to convert a postscript file to a gif. I was trying to convert postscript graphs to gifs because I was producing graphs with 200,000 or more data points, and these graphs were so big that they took a really long time to print if they did print; I was also putting the graphs into a latex file and having 5 of these graphs in the file meant I could never print the dvi version and I had a hard time even loading the dvi version because just clicking on the window would mean all these graphs would have to be reloaded, which took a really long time.

Click here for data.


Smith CS Page / Brown CS Page / Vicky's Homepage / Amy's Homepage / Weekly Journal