Maxwell's Equations --> Dimensional Analysis --> Distance --> Kevin Bacon

Tesla makes this mistake with their ambiguous "lock" button on the iOS app. The button toggles between an open/closed padlock icon. I often lock the doors remotely and since I can't physically see the car to confirm, I get paranoid about whether the icon is representing state or action.

D-sharp and E-flat are two different notes used to describe the same physical vibration for the same reason "father" and "son" are two different words that could describe the same person. It's just a way to communicate contextual relation.

This is false. D-sharp and E-flat have (slightly) different frequencies. Read the article!

...in certain temperaments, but not the ubiquitous 12-TET.

... as thoroughly explained in the article.

Well I haven't read the article, so IDK

I do something similar but I include informational notes as well (not just action items/todo's).

  * incomplete todo

  ~ completed todo

  - informational

Each day I create a new note, titled by date. Grep the folder for topics or incomplete action items. No tooling or overhead needed. I have never felt the need to syntax "ongoing" or "obsolete" items. It's either open or closed.

I found a "normal mode" solution in a couple minutes after some trial/error. But then I spent 10 minutes trying to find a solution in "hard mode" to no avail. I ultimately gave up and wrote a recursive solver in python. Now I feel better about myself and can continue about my day.

This exact product is mentioned in the article as evidence for clinical/market demand. It also shows precedence for FDA approval of a pretty similar instrument for clinical diagnostics. Although, its interesting to note that the hyperfine MRI (at least ostensibly) seems to do the same thing in a smaller/more portable form factor.

I missed it, but I guess my underlying question remains: what is the innovation over previous attempts that gets the paper into nature?

I enjoyed reading that! Great job.

For anyone who found this video satisfying, you may enjoy "Structures : Or Why Things Don't Fall Down" by Gordon.

This plus one thousand. Great book. "The new science of strong materials" is also good. (By the same author)

"Any idiot can build a bridge that stands, but it takes an engineer to build a bridge that barely stands."

A favorite quote from these types of engineering analyses.

Haha, so true. Structures can have multiples of strength for safety margins. Airplanes have 50%. Spacecraft - 10%. There's no way anyone is going to design spacecraft parts without being very good at math.

“furniture designers, incredibly, are not taught during their formal training how to calculate the deflection in an ordinary bookshelf when it is loaded with books,” ― J.E. Gordon, Structures: Or Why Things Don't Fall Down

Author here: I'm equally surprised it worked so well with such little training data. Each image contained 1 example of each character (18 total characters * 10 images = 180 examples). Having said that, I don't think it would generalize well to other people's handwriting until I provided a (lot) more training data.

Did you use any regularization? Did you train until the training accuracy converged?

I feel like the network has enough capacity to overfit 180 training samples

I experimented with a few regularization factors and ultimately settled on a lambda of 0.1. Rather than using a stopping criteria, I ran a fixed number of training iterations (~100) and just eyeballed the cost function results. Since my total training time was fairly brief (~2 minutes, tops), I had the luxury of designing the ANN somewhat heuristically.