Columbia Fall 24 Recap

A few thoughts on the past semester. This is by no means a comprehensive review.

EE 3201 Circuits

Here’s a snippet from our final exam: derive node voltages \(\pmb V_1\) and \(\pmb V_2\).

The answer is:
\(\begin{bmatrix} \frac 1 R + j\omega C_1 + \frac 1 {j\omega L} & -\frac 1 {j\omega L}\\ -\frac 1 {j\omega L} - \frac{\beta}{j\omega L} & \frac 1 {j \omega L} + \frac{\beta}{j\omega L} + j\omega C_2 \end{bmatrix} \begin{bmatrix} \pmb V_1\\ \pmb V_2 \end{bmatrix} = \begin{bmatrix} -\pmb I_{S1} + \frac{\pmb V_S}{R} \\ \pmb I_{S1} + \pmb I_{S2} \end{bmatrix}\)

This looks complicated, but every bit of it makes sense. If after some derivation, any of these does not seem obvious to you, consider dropping the major.

Unfortunately, if actual numbers are given for \(R\), \(L\), \(C\), working out the algebra will be quite tedious. Typically, don’t plug in the numbers until the very last step.

EE 3801 Signals and Systems

Another beautiful course

Exams are really, I mean really, fair game. Questions directly come from the book, lecture, or homework, and there is never anything incredibly difficult. Midterms does have some time pressure (10 minutes/problem), so make sure to practice those key problems. Speed and consistency is the key.

One rant is with exam cheat sheets. The intention is good–reducing our memorization. However, creating a cheat sheet often becomes another form of brute work. You have to admit that copying down every proof instead of trying to fully understand the derivation is easier and faster during preparation, and it’s definitely much faster at the exam. Having more time and harder problems would reduce the reward to such practices. However, given the current stands, although I don’t like cheat sheets, I still take full advantage of it…

Here’s some cheat sheet resources.

• My version for exams 1 and 2 have numerous errors/inaccuracies. It’s hard to be flawless, unfortunately, and I was too lazy to polish over them. I left the old versions in the classic branch

Showing off my final grade there (I actually got 100%, since lowest HW and first two midterms dropped loll)

EE 3106 SSDM

Exam time was really tight, and apparently the prof doesn’t realize that (it’s her second year teaching). Midterm 2 was 15 pages 75 minutes (5 minutes/PAGE!). To that point, you basically need to turn yourself into GNU Make:

1. Find the equation (recipe) of the desired quantity in the equation sheet (Makefile)
2. Recursively make each recipe from its own recipe
3. Write down your make process

Here’s Midterm 2, Problem 1:

Let’s do zero bias: here’s the Makefile (aka equation sheet)

W: epsilon, V0, q, Na, Nd
    W = sqrt(2*epsilon*V0/q * (1/Na + 1/Nd))
Vo: kT/q, Na, Nd, ni
    Vo = kT/q * ln(Na*Nd / ni^2)
kT/q: k, T, q
    kT/q = k * T / q
epsilon: eSi, e0
    epsilon = eSi * e0

Here’s your current directory (aka constants sheet and problem statement)

$ ls -l
Makefile
q       1.6E-19
kT/q    0.026
e0      8.85E-14
eSi     11.8
ni      1.5E10
Na      2E19
Nd      8E15

and you make W!

kT/q = k * T / q
Vo = kT/q * ln(Na * Nd / ni^2)
epsilon = eSi * e0
W = sqrt(2 * epsilon * V0 / q * (1/Na + 1/Nd))

And here was what GNU Ming put down to his stdout:

CS 3157 AP

You’ve probably heard of some horror stories of this class

Yes, after surviving an intensive 15-week brain workout, I can confirm the horror, but it’s definitely manageable with the correct approach.

Students are denied exceptions for the sake of fairness to the whole class

Yes, cheaters outscoring me on homework is infuriating!

This stuff is not easy, even for those who like it

The class is basically 3 parts:

Material

1. C programs
   Theme: pointers and applications
   • E. array, heap, string, struct, function
2. Multiple programs
   Now let’s move beyond a single program, and learn how programs interact with others.
   • Libraries
   • Standard and file IO
   • Multiprocessing: fork() and execl()
   • UNIX: IO redirection, shell scripts
3. The internet
   Now let’s move beyond our local machine
   • TCP/IP: netcat
   • HTTP

and that’s it! There’s nothing too much to remember

Labs

Workload can be a bit lighter, or could be a lot heavier

I can confirm that the labs are pretty…easy, with clear instructions, expectations, and hints.

Most problems are fair-game coding problems. Not much Data Structures background needed. You don’t need to be a competitive programmer.

Do start the labs EARLY. It’s not that they would take the full week to complete. However, it’s important to reinforce your understanding. This is a tight class, and knowledge does build up.

Exams

Exams may not be super accurate in terms of assessing your understanding. That’s a trade-off that I have to make.

Yes, Jae admits this in the first day. Yet it takes a semester for me to understand.

• Exam 1 appears tricky, but if you know what you’re doing, and what the programs are doing, you’ll be fine. My approach was to directly tackle the programs with the memory diagram.

• Exam 2 may have open-ended coding problems, which got me. There are a few tricky details you need to keep in mind with coding from scratch. Still, don’t panic. Approach them in a similar way.

• Exam 3 is focused on multiprocessing and the internet, which actually makes it easier to prepare. No need to remember to boiler plate, but be sure to remember the structure of all Jae’s codes.

Generally, there are two types of coding problems

1. Write the code

   This is relatively straightforward. Fill-in-the-blanks problems are much easier than open-ended coding problems, as there is more room to forget details (E. initializing buffer)

2. Predict the output

   I don’t quite like this, because it’s very easy to miss a line of the code, or forget a particular behavior of a function. We are taught to be more skilled in programming, that compiling compiling. This should be gcc’s job.

   • Problems vary in difficulty. Some have useful hints and checkpoints. Some have generous partial credit, while others are “do-or-die”. Those are just due to the nature of the problems.

You may not do well even if you try

Claim: Any grade below an A- is easily achievable via hard and smart work. Everything above may not.

“Proof:” here

It’s not that the problems are impossible–everyone is able to get an 100 on any exam. Rather, it’s the excessive stress and the meticulous attention to detail.

From my personal perspective, getting a decent (90+) exam grade often comes down to a bit of luck. Everyone’s intelligent and well-prepared. It’s just a matter of whether or not you catch your mistake, and if you don’t, you may be very heavily penalized.

Part of your consistency can be improved by repeated practices, but there is always something that is beyond your control.

At the end of the day, we are human, not machines, not computers, not gcc -g -Wall. We make mistakes, and that’s part of what makes us human.

The takeaway is: don’t over-stress about your AP grade, and don’t let Jae Lee hurt your confidence in your ability, intelligence, carefulness, or hard work. As he said, in the end, this is just another course, and you have other priorities in life.

Some people think I’m a great professor. Some people think I’m a horrible human being.

In fact, this course is perfect in every aspect: the instructor, TAs, CLAC, lectures, labs, exams, and grading. You may not get the result you want, but you can’t say anything wrong about its design and content. It is that good. I thoroughly enjoyed a dense portion of it–the only exceptions being the three (actually 6) instants when exam solutions/scores are released.

This isn’t meant to be a survivor’s tale. Whether I get an A+ or end up with a C, my opinion remains unchanged: CS 3157 AP is an awesome class!