There are many questions asked in the webboard about these
classes. So I guess I should just sum them up in one place.
Some of these questions are not really specific to the department.
For EE, Electrical Engineering, there are these questions:
I will study in EE in this Fall 2002. Because there are
many 400-level courses that are prerequisites for 500-level
courses, I don't know which classes
I should take. In Fall, these 400 classes...
1. EE 483: Introduction to Digital Signal
Processing
2. EE 241: Applied Linear Algebra for Engineering
3. EE 464: Probability Theory for Engineers
4. EE 401: Transform Theory for Engineers
Are they important or mandatory?
P' Wit:
You can take up to 3 400-level classes for your MSEE. I
took DSP with Dr Leahy, Prob with Dr Welch and I read Linear
Algebra by myself for 2 weeks before I took Random Process
with Dr. Zhang.
In my opinion about 500-level classes that I took in the
SIP track, I rank them as below:
1. EE 569: Introduction to Digital Image
Processing
2. EE 559: Mathematical Pattern Recognition
3. EE 583: Adaptive Signal Processing
4. EE 596: Wavelets
5. EE 562ab: Random Processes in Engineering
6. EE 563: Estimation Theory
7. EE ???: Optimization Theory
8. EE 500: Neural and Fuzzy Systems (Mendel's).
Have fun!
P' Peach:
Check or if you are in LA, get the Catalogue.
It is the most authoritative and supersedes all other document.
1. Of course there are prerequisite. But if you have sufficient
background and talk right, most of the professors who will
be signing your study plan will sign and waive the 400 course
for you. For example, most students with Computer Science
degree get waived for CS402 Operating System (required in
MSEE Computer Network) but a student with BSEE will probably
have to take it unless he can convince the professor that
he knows enough of Operating System.
2. Check degree requirement for your program. For example,
MSEE Computer Network requires you to take either EE 550
or EE 555. This should answer your question about which
course is mandatory.
3. Plan ahead. Plan what classes you are taking for all
three or four semesters you will be here. Some classes are
not available in Summer. Some only available in Spring,
others in Fall. Some require prerequisite. Some are so popular
that you have to stand in line from 6 am to get the D-Clearance
for registration. If you don't plan carefully, you may end
up spending an extra semester just because the last required
course is not available in the semester you intended to
be the last.
4. Talk around a lot. Know how the course is like. You don't
want to be the only Master student in a class that is taken
mostly by Doctor students. It's wise to stick with class
that has a few Thai student. Study at USC is *HARD*. Lots
of work. You will want to talk to someone in the same class.
P' Ake:
I personally recommend any DSP or COMM student to take either
Prob+Linear Algebra+Transform or Prob+Linear Algebra+DSP.
Except someone who really good like Ake(Krisda), Golf, or
Kriang (All have already graduated). This is because these
3 classes are basics for everything in Communication & Signal
Processing. You will never understand anything in very deep
sense if you're not proficient in these topics.
1. EE 464: Probability Theory for Engineers, this teaches
you how to characterize the relationship among variables
in the system and how to get the desired statistical quality
and quantities out of it. In Prob., you will learn a simple
system with one or two variables. "Assuming your professor
is good", you will learn
- The Classic Probability Theory: including some combinatorial
and counting methods
- The Modern Axiomatic Probability Theory
- Random Variable with its characteristic: => density/distribution
function, moment, transformation of random variable etc.
- Two Random Variables with its characteristic: => everything
in jointly density/distribution function, moment, transformation,
and "estimation" (edif your professor want to include)
- Some more topics like Central-Limit-Theorem, Weak/Strong
Law of Large Number
- etc. (e.g., I learned some quantization, radar detection,
etc. in this class too)
- Oh...you might learn some "random vector" too.
The bad thing about this class is that the quality is very
very different from one professor to another. Then, the
larger system will be taught in EE562a (random processes)
which extra tools from Linear Algebra and Transform Theory
is required to characterize these more-complicated system.
Let say...Random processes "at USC" (not other school though)
start with
- Hilbert Space, Metric Space, Norm Space, ...
- Random Vector with all Linear Algebra (the second part
of Linear Algebra class) immediately such as KL theorem,
Mercer Theorem, Eigen value/vector, etc + all finite-length
discrete-time transform from Transform Theory => Discrete
Fourier Transform
2. EE 401: Transform Theory for Engineers, it teaches you
how to change the system to different domain for both discrete-
and continuous- time. Basically, you will learn
- Complex number theory (I forget the correct name) including
the residue theorem => This complex thing is very important
for all types of transform.
- Fourier Series (for periodic continuous-time signal),
- Fourier Transform (for non-periodic/periodic continuous-time
signal),
- Laplace Transform (generalization of Fourier Transform)
- Discrete-Time Fourier Transform (for discrete-time signal)
- Discrete Fourier Transform (for finite-length discrete-time
signal)
- Z Transform (generalization of Discrete-Time Fourier Transform)
3. EE 241: Applied Linear Algebra for Engineering, this
class is extremely important and what I would like to say
is that the topic is not as simple as what people might
think (i.e., solve Ax=b). This class can be divided in 2
parts (after some introduction for Ax=b using Gaussian elimination)
Part 1: Space Theorem
Row/Column Space, Dual Space, Inner Product, Rank, Dimensionality,
Basics,
Finite Field (Galois Theory) -- Prof. Golomb likes to use
without telling you he used it!, Orthogonality Principle,
etc.
Part 2: Eigen Theorem
Eigen Value/Vector KL & Mercer Theorem Determinant Diagonalization,
etc.
Let me show you some examples how important the Linear algebra
is (Prob & Transform is very very important too but...I
talk only Linear Algebra since someone say it is easy).
1. All "structure in probability" can be viewed from "algebraic"
point-of-view too. And algebraic point-of-view is a part
of Linear Algebra. e.g., correlation between 2 variables
in the system can be measured with inner-product operation
(Linear Algebra concept). Also, ortogonality + independent
in probability is just another view of orthogonality in
linear algebra.
2. All Part 1 of Linear Algebra is a basic of error correcting
code. Typically each codeword is just an element in the
"Subspace" and each code can be defined by subspace.
3. The transmission in communications is nothing but transmitting
a vector of data to the channel that can be characterized
with its Eigen function, If channel is color with null exigent
value in some exigent direction, the transmitted data should
be represented with vector along that exigent function.
Don't think (3.) is too much....you will get (3.) in Random
Processes (EE562a) "before midterm". That is why random
processes at USC is quite tough. It is a generalization
+ combination of Prob.+Linear+Transform. If you don't have
enough skill in these 3 topics (need a lot of hard works),
you will never understand why USC teach Random Processes
like this.
If you would like to know whether your skill is good enough
without taking class. Traditionally, EE562a has a "test
homework" in the first class that measures you whether you
are ready. I recommend you to get this 1st EE562a homework
and check your skill.
Finally, I don't know much about DSP ^^. OK...last...I may
scare you a little bit (a lot?)....sorry about that.
-- For more details, check out our webboard topic# 41
