As I approach the last three weeks, my tasks have definitely become more writeup-focused. I’m still working with the HPC to test the effects of a greater batch size, and I’m also refining the metrics I’m using to evaluate my model. There are two ways to evaluate my work, its raw accuracy and general comprehensibility, and the second is definitely a lot more difficult to translate to technical terms. As the results of my work need to be made clear to non-Persian speakers, this adds another layer of difficulty to the process of deciding which metrics are most appropriate.

This week, one of my friends from my exchange was in the city to see family, so we went to the Düsseldorfer Kirmes, the local fair, which was right on the river and packed. German town fairs seem to have a wider variety of foods available, and many different places to get beer as well, which was quite interesting.

On Saturday, I finally took the trip to the relatively nearby city of Arnhem, where I played the carillon located in the Eusebius Church. The church had a very cool all-glass outlook that you could step onto, and it had a great view of the city.