Lessons from Optics, The Other Deep Learning – arg min blog


37 bookmarks. First posted by Vaguery january 2018.


It would be easier to design deep nets if we could talk about the action of each of its layers the way we talk about the action of an optical element on the light that passes through it.

We talk about convolutional layers as running matched filters against their inputs, and the subsequent nonlinearities as pooling. This is a relatively low-level description, akin to describing the action of a lens in terms of Maxwell’s equations.

Maybe there are higher level abstractions to rely on, in terms of a quantity that is modified as it passes through the layers of a net, akin to the action of lens in terms of how it bends rays.

And it would be nice if this abstraction were quantitative so you could plug numbers into a formula to run back-of-the-envelope analyses to help you design your network.

It would be easier to design deep nets if we could talk about the action of each of its layers the way we talk about the action of an optical element on the light that passes through it.

We talk about convolutional layers as running matched filters against their inputs, and the subsequent nonlinearities as pooling. This is a relatively low-level description, akin to describing the action of a lens in terms of Maxwell’s equations.

Maybe there are higher level abstractions to rely on, in terms of a quantity that is modified as it passes through the layers of a net, akin to the action of lens in terms of how it bends rays.

And it would be nice if this abstraction were quantitative so you could plug numbers into a formula to run back-of-the-envelope analyses to help you design your network.

There’s a mass influx of newcomers to our field and we’re equipping them with little more than folklore and pre-trained deep nets, then asking them to innovate. We can barely agree on the phenomena that we should be explaining away. I think we’re far from teaching this stuff in high schools.
deeplearning  optics  physics  education 
20 days ago by mike
Trying to frame our discussions of deep learning science, when we are still pre-newtonian in a lot of it...
deep  machine  learning  abstraction  mental  model  design  science  epistemology 
april 2018 by asteroza
How deep learning could to be more like optics
physics  deep-learning  science 
february 2018 by pmigdal
Favorite tweet: deliprao

I love this analogy of #DeepLearning to optics by @alirahimi0. There is a lot we don't understand. That's okay and not okay. Differentiable programming enabled by hardware advances is here to stay. Let's continue to make sense of it! https://t.co/kRmH0Ntc7f pic.twitter.com/ZJxZ3IPsrO

— Delip Rao (@deliprao) February 12, 2018

http://twitter.com/deliprao/status/963155300927750144
IFTTT  twitter  favorite 
february 2018 by tswaterman
An excellent post describing how I have been feeling about DL methods:
ATM it's hard to mov…
from twitter_favs
february 2018 by kartik
Lessons from Optics, The Other Deep Learning
from twitter_favs
february 2018 by danbri
Imagine you’re an engineer, you’re given this net, and you’re asked to make it work better on a dataset. You might presume each of these layers is there for a reason. But as a field, we don’t yet have a common language to express these reasons. The way we teach deep learning is very different from the way we teach other technical disciplines.

A few years ago, I got into optics. In optics, you also build stacks of components that process inputs. Here’s a camera lens.
machine-learning  philosophy-of-engineering  system-of-professions  pedagogy  knowing-what-you-know  epistemology  engineering-criticism 
january 2018 by Vaguery