The Black Box that is Deep Learning

-
Deep Learning has — fundamentally — changed so much about how we approach problems across a ridiculous number of areas, and it’s only accelerating. It’s not just about doing amazing things with images like Style Transfer, or Painting, or even removing Henry Cavill’s mustache, it’s now getting deeply embedded in tools for Marketing, Dentistry, and Voice synthesis (because you want your GPS directions to be read out by your spouse maybe?).
It’s all really good stuff, with one underlying issue, we don’t really know how the stuff works!
Take the de-noising of images for example. Before Deep Learning came around, there was a ton of work — all rigorously based on mathematics and science — that went into the algorithms that were used. Partial differential equations, anisotropic diffusion, and a host of other techniques, all of which have solid theoretical foundations.
And then, in 2012, Burger et al. threw Deep Learning at the problem, and basically beat the pants off of the state-of-the-art in de-noising at the time .
The kicker here, of course, is that with Deep Learning we have a very limited understanding of the what actually goes on under the hood. Oh, we’re starting to formalize some of this (latent representations, etc.), but most of it still boils down to some combination of
1) Get a ton of data
2) Throw a whole bunch of GPUs/TPUs with TensorFlow/Keras/… at it
3) Profit
Operating in this black-box style has consequences, with the major one being that they make the process of moving from Data → Information → Knowledge that much harder. The lack of interpretability and our (current?) inability to understand the underlying models limit the knowledge we can glean from these systems. Mind you, the results are amazing, it’s just that it would be useful for them to be understandably amazing (Scientific Method FTW!)
Mind you, it’s not that bad — after all, we don’t quite understand how the human mind works either, so there you go . That said, I’m with Michael Elad, when he says we should “allow deep learning to influence [his] research team’s thoughts and actions, but … continue … seeking mathematical elegance and a clear understanding of the ideas [they] develop”
(•) “Image denoising: Can plain Neural Networks compete with BM3D?” — by Burger et al.

Comments

Post a Comment

Popular posts from this blog

Erlang, Binaries, and Garbage Collection (Sigh)

-
Image
The nice thing about garbage-collection is that it is makes memory management absolutely transparent to you Of course, the above is true right up to the point when it stops being transparent, at which point it does a phase change from nice to clusterf**k of biblical proportions .  There are entire posts, books, and suicide notes that have been written about this, and I have no great desire to add to them, save with the minor comment that paranoia about garbage-collection is still justified , albeit only in the edge-cases for most people. That said, I'll give you that Erlang GC is in many ways much, much better positioned in this regard than, oh, Java, but it still rears up and bites you in the butt when you least expect it. ( More information about Erlang's GC vis-a-vis Java here , and Jesper's truly excellent description of Erlang's GC here ).  A story from the past should serve to illustrate this. Once upon a time, in a magical land far far away - well, Her...
Read more

Spot The Difference — Deep Learning Edition

-
Image
You’ve played  Spot The Difference  at some point, right? (Of course you did. We  all  did  ). Think about what you actually did when you were trying to Spot The Difference — in fact, just do it again with SpongeBob juggling above. You’re probably doing something like 1. Quick glance to the left. 2. Quick glance to the right 3. “ Spatula on the left. Wonder if there is one on the right? ” 4. “ Yup, there is. Sriracha on the right. Is that on the left? ” 5. “ Nope! It’s a ketchup bottle on the left! And there is nothing on the plate below it! ” etc. To summarize, what you’re basically doing is looking at each image alternately, identifying objects in each image, and trying to find the same object in the other image. As it turns out, what most Deep Learning systems do is something quite different. They basically do something like this 1. Figure out everything in the first image (“ ketchup bottle at 1 o-clock, empty plate at 3 o-clock, spatula at 1...
Read more