Hacker News new | past | comments | ask | show | jobs | submit login
Help Advance the World with Advanced Linear Algebra (utexas.edu)
109 points by MaysonL 60 days ago | hide | past | web | favorite | 32 comments



Why EdX and not OpenCourseWare (OCW)? Before Coursera, a lot was freely available without having to go through the agreements and overhead required to access via a MOOC like EdX.

Harvard and UC Berkeley used to have an excellent OCW sites but they migrated a lot of content to EdX or elsewhere and removed it from their OCW sites. Stanford also took a lot of stuff off of their multiple OCW type of sites and moved it to Coursera or elsewhere. Thankfully, MIT still kept much of their stuff on their OCW site even after adding it to EdX.

Ten to fifteen years ago a lot of universities had sites imitating MIT OCW but now very few do. The trend has been to move very good freely available notes, etc., off of faculty webpages to behind Blackboard or similar logins; now there is little left that is publicly available and easily accessible without having to login to some site (or be tracked).

I spent a lot of time searching for reasons why universities went away from OCW. Apparently, people with hearing disabilities sued universities to get transcripts added to OCW courses that had videos on the grounds that lack of transcripts left them at a competitive disadvantage. This didn't make sense to me: many OCW courses had great course materials but no videos; courses with videos could simply release one video at a time with the next video held off until the general public / free users / people like me completed a transcript on a wiki type site of the previous lecture. (For some of those course materials, I would have gladly worked on transcripts on a wiki for free!)

In case there is anyone out there that isn't aware of any of these:

* https://www.oeconsortium.org/

* https://ocw.mit.edu/index.htm

* http://ocw.jhsph.edu/

* ...


I had no idea they were doing this stuff to material I've benefited from accessing. Thanks for taking the time to give us the benefit of your research.

Control freaks always want your personally identifying data, and people are too open to giving it out. HN is one of the greats because they are not such control freaks.


Agreed. Even the class syllabi that used to be published on the public web are increasingly requiring a school login.


I have always been pleased with how willing the mathematics community is to educate others freely.

Some other free Linear Algebra courses:

https://ocw.mit.edu/courses/mathematics/18-06-linear-algebra...

https://open.math.uwaterloo.ca/


Gilbert Strang's Linear Algebra course is often recommended, but I also recommend a set of lectures on YouTube from Technion with Aviv Censor: https://www.youtube.com/watch?v=aefKXYYXT6I&list=PLW3u28VuDA... since it's very beginner friendly and there's a lot of step-by-step explanation.


Thank you. The link in the top-level post isn't educating the world freely. It's a scammy edX course. Yours seem like the real deal.

There's a few more out there too.


I have a bit of a problem with this. I am not sure why it does not sit right with me.

The college needs $75k of donations to be able to put the curriculum on edX?


Watch the video. The instructors want to put it online for free—not the university.

They have already made courses and made them available online. It was a massive undertaking. It is not too much to ask for their effort to be funded at the modest level they have requested.


No. If the instructors wanted to put this up for free, they could build the course on Open edX, and put it up for free with any of the hundreds of free Open edX providers. They could license under an open license. There's nothing stopping them (including their university).

They're asking us to crowd-funded a crappy profit-making scheme.

The course will be of questionable quality. You don't learn math without doing math. That's a scientific fact. Their "free" offering is "audit videos." I'd encourage anyone to look up the mass of literature about interactive engagement versus passive learning. You feel like you're learning, but you're not. People basically would waste their time.


I think there's an argument to be made that a public university should be putting content like this online for free anyway.


You are free to make that argument, and you are free to change the system.

That said, it is counterproductive to make that case in the instance where someone is trying to use public money to provide a public good. The correct place to protest is when the university sends you your tuition bill or when they decline you admission. The instructors have no access to or control over the billions in their university’s endowment fund.

Right now, our options are to either fund this effort or not to fund it. I hope that we fund it.


> The correct place to protest is when the university sends you your tuition bill or when they decline you admission.

Your "protest" over a (1) tuition bill or (2) admissions rejection is not going to matter. You'll be sent to collections in the case of (1) or ignored in the case of (2).

Publicly funded universities should contribute _something_ back to the public. Generally it's fucking football or basketball (assuming you pay for the right TV service).

You are barking up the wrong tree.


> You are barking up the wrong tree.

:)


Yes. We should have high-quality videos for the entire STEM and med school curriculm.

It wouldn't cost that much as a % of public education costs.

Khan Academy claims to be doing this, but their videos are not particularly high quality.


There is, and it might even win eventually, but students could use the material now.


How long does it take to produce a course? Probably 4:1 hours prepared per hour of content, with 33h of material in undergrad courses. Say 200 hours of high quality focus?

Let’s add 80 hours of technical nuisance to make the videos work nicely too.

That means that the effort would be worth $268/h roughly. If the course was very good, I’m not against a teacher making that for a particular project.

Calling it donations is weird though, I guess another commenter said it was the university that was requiring it


I would be more inclined to donate if it was not associated with the University, and was simply the professors creating the content offering.

The amount of money floating around in higher ed., it seems in poor taste to need a ‘donation’ to make this available to the public at large, from a public university, funded by taxpayers.

It could be marketed better.


Also edX now requires payment for the ability to submit and check homework.


Something not touched on in the video or description but I think is worthwhile.

The greatest failing in my math education is that it didn't give me the "keys to the kingdom" in other words, it didn't teach me how to conceptualize problems and really dig into literature to further my understanding on topics as they come up.

Contrast that with my signal processing courses, which were very much focused on concepts, conventions, notation, and best practice to provide a ground work for future study.

And working in the signal processing field, it was easy for me to dig into literature as I researched problems up until my math skills failed and I had to teach myself a lot of what my grades showed I had learned in college. Linear algebra in particular.

So when I see something like a course that will teach me about orthagonality, linear systems, SVD and eigenvalues I want to back it. But only if it's the kind of content that will tell me what those things are and how to read about them and understand them, not more rote memorization of how to work a problem set.

Just my two cents.


I recall covering these topics in my undergrad linear algebra course. Textbook was: https://www.springer.com/gp/book/9780387940991

Probably not as deeply as this course proposes to be sure, but good into for those interested.


It's standard LA fare... Pretty much any LA textbook will cover these. For a focus actual computation (ie numerical LA), consider Golub's (RIP) Matrix Computations (more of a reference, though, not so much for self-study), Trefethen's Numerical Linear Algebra, Demmel's Applied Numerical Linear Algebra.


How is this "advanced" linear algebra? This was all covered my freshman linear algebra course.


It's the same topics as first-year linear algebra, but now done "for real" using algorithms that scale.

Part I: SVD = bread and butter in industry with lots of applications in ML and stats, engineering stuff too, see https://www.youtube.com/watch?v=R9UoFyqJca8 )

Part II: Solving Linear Systems = 70% of science can be describes as A*x=b where A is a matrix and x and b are vectors. Spoiler alert: the "find the RREF algorithm" that we learn in first-year LA is not the most efficient (or numerically stable) option.

Part III: Eigenvalues "for real" this time = e.g. iterative solution methods that scale. The flagship application of this idea would be the initial PageRank algorithm that works on the adjacecy matrix representaiton of the graph of all webpages (millions of columns, see http://infolab.stanford.edu/pub/papers/google.pdf )

So to summarize, this course covers all kinds of topics that you don't need to know if you want to apply LA to small and medium data, but for large scale stuff would be really good to know.


I don't think that it is helpful to classify basic vs advanced linear algebra according to whether the techniques "scale" to larger data sets. Linear algebra isn't necessarily about data sets at all. Many of these courses (even this "advanced" one it looks like) completely fail to convey the understanding that vectors and linear transformations on finite-dimensional vector spaces only acquire numerical representations as numerical vectors and matrices when a basis is specified; but that should be part of a basic linear algebra course.


I take it they will go deeper into the materials than they went in their previous course (http://www.ulaff.net), which also covered the topics to a limited extend.


Don't know why you are downvoted but I second your opinion. The contents listed here seem to emphasize applications in CS rather than theoretical aspects (e.g., Hilbert space, tensors, spectral theory).


"Applied" or "Numerical" LA would be better descriptors, yes.


ime most introductory linear algebra classes do not affect students to develop a robust mental model of the fundamental theorem of linear algebra. the first picture in the course description is this very model, so ideally this class does that. i hope...


Can anyone here say anything about the quality of this UT Austin online CS masters program? (The online masters programs I've looked at previously I haven't been very impressed by.)


Based on best preliminary evidence, the Georgia Tech one is dramatically better. A lot of shitty stuff (like this) has been coming out of UT. Georgia Tech seems to have made a serious investment in doing online right.


I don't have an answer to your question.

Since they haven't started the course yet, I doubt anyone has direct experience,

but the curriculum for the fall 2019 fall offering is (from their home page [1] )

1. CS 391L Machine Learning Adam Klivans and Qiang Liu

Computing systems that automatically improve their performance with experience, including various approaches to inductive classification such as version space, decision tree, rule-based, neural network, Bayesian, and instance-based methods; as well as computational learning theory, explanation-based learning, and knowledge refinement.

2. CS 380L Advanced Operating Systems -Vijay Chidambaram

Study of the formal structure, design principles, organization, implementation, and performance analysis of multiprogramming and/or multiprocessor computer systems.

3. CS 380P Parallel Systems - Calvin Lin and Chris Rossbach

The objective of this course is to provide students with strong background on concurrency fundamentals along with experience with a diversity of both classical and modern approaches to managing and exploiting concurrency, including shared memory synchronization, parallel architectures such as GPUs, as well as distributed parallel frameworks such as MPI and map-reduce. Material will be presented through readings and discussion of background material as well as occasional recent research papers when appropriate. The course requires a number of programming and project assignments to provide direct experience with design, programming, and measurement methodologies for concurrent systems.

4. CS 395T Optimization -Sujay Sanghavi

This class covers Linear Programming and Convex Optimization. These are fundamental conceptual and algorithmic building blocks for applications across science and engineering. Indeed any time a problem can be cast as one of maximizing / minimizing and objective subject to constraints, the next step is to use a method from linear or convex optimization. Covered topics include formulation and geometry of LPs, duality and min-max, primal and dual algorithms for solving LPs, Second-order cone programming (SOCP) and semidefinite programming (SDP), unconstrained convex optimization and its algorithms: gradient descent and the newton method, constrained convex optimization, duality, variants of gradient descent (stochastic, subgradient etc.) and their rates of convergence, momentum methods.

5. CS 395T Deep learning -Philipp Krähenbühl

This class covers advanced topics in deep learning, ranging from optimization to computer vision, computer graphics and unsupervised feature learning, and touches on deep language models, as well as deep learning for games.

(Spring 2020)

6. CS 395T Advanced Linear Algebra for Computation Maggie Myers & Robert van De Geijn

Linear algebra invariably lies at the core of techniques that are of critical importance to computational and data scientists. In this course, you learn advanced concepts in linear algebra, practical algorithms for matrix computations, and how floating point arithmetic as performed by computers affects correctness.

7. CS 388G Algorithms: Techniques and Theory Vijaya Ramachandran and Greg Plaxton

Sorting and searching algorithms, graph algorithms, algorithm design techniques, lower bound theory, fast Fourier transforms, NP-completeness

8. CS 395T Online Learning and Optimization Constantine Caramanis and Sanjay Shakkottai This class has two major themes: algorithms for convex optimization and algorithms for online learning. The first part of the course will focus on algorithms for large scale convex optimization. A particular focus of this development will be for problems in Machine Learning, and this will be emphasized in the lectures, as well as in the problem sets. The second half of the course will then turn to applications of these ideas to online learning.

9. CS 394R Reinforcement Learning: Theory and Practice Peter Stone and Scott Niekum Introduces the theory and practice of modern reinforcement learning. Reinforcement learning problems involve learning what to do—how to map situations to actions—so as to maximize a numerical reward signal. The course will cover model-free and model-based reinforcement learning methods, especially those based on temporal difference learning and policy gradient algorithms.

assuming these are the equivalent of the offline courses, in terms of instruction, exams/assignments, and have decent instructor access, if not at quite the level of an on campus education, looks solid enough to me.

[1] https://www.cs.utexas.edu/msonline


I like the Georgia Tech's approach to this. Most of their courses from the online master's are also available freely as mooc on udacity.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: