I thought this semester was difficult.
I appreciate the knowledge I gained from this course. I learned how to code (with PyTorch) and train various neural architectures from scratch, starting from a basic MLP to more complex models such as CNNs and Transformers. However, this course took too much time for what I have gained from it. Assignments can feel like pay-to-win, as some models take a very long time to converge (I think my HW2 model took 50+ hours to converge), so it is incentivized to deploy (larger) models on higher cost GPU instances (A100s, etc.). Some advice to succeed is to go to office hours and don’t do anything too crazy with your models. Look at reference SOTA papers if you are feeling lost.
This class is a standard systems class. It covers the network stack and common associated protocols, with an emphasis on higher levels of the stack (think layers 4-7). Projects are in C (despite the C++ infrastructure), which can be a pain to work with as many labs require data structures which are not in C’s standard library. The first project is a mock link-layer protocol, using OSPF for the data plane, STP for control plane, and some additional stuff (mixing, random routing) for security. The second project is building TCP on top of UDP, as well as an associated congestion control algorithm based off of Reno. This was probably the hardest lab. The third project is a generic HTTP 1.1 server and client. Overall, the course did a very good job introducing networking concepts.
I don’t have much to say about this class. First third of the course covers very standard counting techniques such as PIE and probabalistic method. We then went into notable topics in combinatorics such as extremal combo, linear-algebraic combo, and Ramsey Theory which I found somewhat interesting. However, I felt that the topics in this class felt sparse.
Don’t be fooled by the name of this class; I found it not to be as interesting as I thought. This course covers various aspects of human aptitude such as its measurement and its underlying factors. Grades are determined through writing assignments and a final.