Home Creators Posts Import Register Favorites Logout
haven't archived this post yet. have a subscription? use the importer!

Content

Hi Everyone,

I mentioned in a previous post that I'd be collaborating with an MIT group on a course they're running about "computational thinking" this semester.  Well, it's happening!  Linked above is the first live lecture I gave.  There was also an asynchronous lecture earlier in the week giving the basics of working with images in the language used for this class (Julia).

Course intro: https://youtu.be/vxjRWtWoD_w

Lecture 1: https://youtu.be/8rrHTtUzyZA

It's a funny little elective course, where the plan is, admittedly, rather scattered.  There are four units, image processing, ray tracing, epidemic modeling, and applications to climate science.  Each one is meant to give a context where we can talk about some general principles in computation and math.  Also, given that the group running the course is MIT's Julia lab, it's effectively a course on working with Julia as much as anything else.

My own role here is simply to give the occasional lecture and generally provide advising on the nature of teaching over the internet.  So far, it's been a wonderful chance to learn new things and play around with different styles, e.g. putting the code much more front-and-center than it typically is in a lot of my other work.

It's not entirely decided how frequently I'll lecture.  The general goal is to be helpful to the course, but without interfering too much with keeping things moving on the 3blue1brown front.

------

Also, the Hamming code videos are now out, including Ben Eater's, which you should most certainly check out.

Ben Eater's video:
https://youtu.be/h0jloehRKas

Part 1:
https://youtu.be/X8jsijhllIA

Part 2:
https://youtu.be/b3NxrZOu_CE


I hope you're all doing well,
-Grant

Files

Previews only

Convolutions in image processing | Week 1 | MIT 18.S191 Fall 2020

The basics of convolutions in the context of image processing. Course website: https://mitmath.github.io/18S191/Fall20/ Join the discussion at https://discord.gg/RAyDRr Contents: 0:00 - Introduction 1:10 - Simple box blur 4:31 - Gaussian blur 7:26 - Looking at the convolution in Julia 13:00 - Example of a non-blurring kernel 17:13 - Sobel filters 21:25 - Relation to polynomial multiplication 26:08 - Relation to Fourier transforms 34:00 - Final thoughts

Comments

Anonymous

Congratulations!

Gabe

Looked good!

Anonymous

Hi Grant, do you also do a lot of programming in Julia? One of the things I liked most about the language (apart from the dot and nerdy things like metaprogramming) is the libraries for automatic differentiation, especially Zygote. It's amazing to me that you can just write an complex algorithm as you would program it and then differentiate it. Maybe the principles behind this would be a cool video? I read some stuff from Zygote about adjoints and vector-Jacobian products but even though I know quite a bit of calculus it was all very unfamiliar to me ;)

Anonymous

As someone who needs this kind of understanding "only sometimes" as a professional (my last linear algebra or calculus class was over 25 years ago), this was a really helpful reminder of what convolving really means intuitively. But I think your comment is really right: 30 minutes is too fast and clipped. Your style could easily spread out over 45 minutes without dragging. I wish you could have just a little more time to not feel rushed.

Anonymous

Great, clear, and filled with new learnings that I always enjoy in your videos!

Anonymous

Thank you for this, I'm excited to check these out!

Doug Fort

This is great. I'm going to try to take the course like a real student.

Anonymous

My daughter is named Julia so I approve.

Anonymous

The pacing might need to be different for a lecture versus a video. I notice my own kids play most content at 2X speed.

Anonymous

I hope the notebook will be shared so that we can run the examples and extend them

white beard geek

Enjoyable and clearly explained. I am looking forward to more.

3blue1brown

I don't do a ton of Julia programming, but you're completely right that automatic differentiation would make a good topic, and that this language makes it particularly natural.

John Nichols

Can you share what tools you used for the convolution stepping visualization and the polynomial multiplication part?