Losses, gradients, backprop, dropout? Sorry, but you might as well bring up kernel functions or hinge loss
Show this thread
People constantly try to interpret what the mind is doing using analogies ("memory is like a filing cabinet..."). But my least favorite kind of analogy is when deep learning people compare properties of the mind to ad-hoc deep learning concepts. You people should know better
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Do you believe that improving our understanding of the mind could lead to improvements of deep learning methods? Or maybe new paths to follow in machine learning?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Completely agree. I’m not sure why that type of analogy is made.
-
We naturally and spontaneously think of other beings — natural and artificial; gods, animals and man-made devices — in anthropomorphic terms.
- Show replies
New conversation -
-
-
Francois, try to address the root cause. People who dont need to learn it are doing so and then people who shouldn't be teaching are teaching and that leads to such consequences. They grasp for what sounds easiest. As ma boi
@nntaleb said: making sense <> being rightThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
A good mental exercise is to compare machine learning algos to some other complex decision making system, for example the immune system, instead of the nervous system.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Filing cabinet is particularly bad, memory runs on something not yet reflected in deep learning
-
Deep learning is like a blind man looking for his keys, it can work to find patterns, but really inefficient, human mind has a whole fundamental system not accounted for in Deep Learning, there are different levels of learning, “Deep” learning is actually quite shallow
End of conversation
New conversation -
-
-
ReLU ==> 1 when my brain attends, 0 when it does not?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
The point of deep learning is to compare it to the brain’s cognitive abilities. The fact that we can’t do it effectively yet doesn’t mean that we shouldn’t examine the similarities and differences between functionalities, actually it means that we should.
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.