It’s not gradient descent unless you send it other people, get feedback and iterate. And even then it’s a very weak analogy. Rewriting a draft is more akin to code refactoring. You realize the intent you initially had and you figure out how to express it better on every edit.
The similarity between writing & training a neural net is striking. Start with random weights (AKA a bad first draft), and apply gradient descent until it plateaus / you can't bear it any more...
-
-
-
For you, maybe. For me, it's much more like gradient descent.
- 1 more reply
New conversation -
-
-
Oscar Wilde said that writing was easy. You just sit at a typewriter and bleed.
-
I forget who else said that it is simultaneously the only thing and last thing you want to do.
End of conversation
New conversation -
-
-
And sometimes it just won't converge...
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
That’s pretty much all iterative processes.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Both are creative/iterative processes that spawn knowledge with increasingly hard to vary explanations and are at the beginning of infinity? Random weights = “conjecture” and gradient descent = “criticism”. So really both can be thought of as science?
@DavidDeutschOxfThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
What I like most about this analogy is the emphasis that the first stage of writing/coding/creating is very often just random initialisation. One of my most treasured realisations - though it took me years...
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Both are fundamentally some optimization process after all ...
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Life is Machine Learning
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.