@fchollet Some brilliant researchers (LeCunn, Hinton) have developed powerful techniques (convs, dropout) by analogy with neural processes.
In "neural networks", the neuro analogies represent an unnecessary and actively harmful level of conceptual indirection.
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@ferrouswheel@fchollet oh sure, but I found Hinton's reasoning about dropout to be incredibly interesting and enlightening.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@ferrouswheel It's useful to remember that they are just analogies though. Neurons in the brain have nothing to do with electronic ones.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@fchollet Agreed but it's fairly common: e.g. labour/republican/etc parties in most countries, that are not that pro-labour/republicanism/…Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@fchollet we should call it the "balls-and-sticks networks". The more balls you have connected with more sticks, the better they learns.Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
@fchollet Indeed. I always flinch when someone emphasises the "like a human brain" aspect of a model, as if that is the key insight -
@fchollet But I think the name is there to stay, and provides a convenient term to group together a number of related ideas and techniques
End of conversation
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.