Comparing optics to deep networks:
"That’s not because optics is easy. It’s because the mental models in optics are organized well.
Modern optics is taught in layers of abstraction."
h/t @aeronlaffere
http://www.argmin.net/2018/01/25/optics/ …
-
-
This goes beyond my knowledge, just a question: would having specific models at various layers of abstraction lead to a specific output at each level?
-
By specific you mean predictable?
-
Yes, at least within a range.
-
If so DL would stop being a black box
-
Mostly agree. I think understanding (i.e., shining a light into a black box) is linked to predicting but not so straightforwardly.
-
Probably not. I would guess that formal analysis is much needed
-
Yeah, exactly. You can not understand something (ergo still a black box) and yet still memorise and even duplicate all the input output mappings.
-
This is why I take a stand when people say cog sci is reverse engineering the brain or mind. It's so much more than that!
- 4 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.