entralprocessing unit (CPU)/graphics processing unit (GPU
I'm interested to know how these constraints relate to TPUs
7 Matching Annotations
- Nov 2023
-
mml-book.github.io mml-book.github.io
-
-
additional term to remember what happened in theprevious iteration
Reminds me of recurrent neural networks/LSTM
-
numerical optimization meth-ods
Reminds me of Numerical Methods for those who took that class here at Guelph
-
- Oct 2023
-
mml-book.github.io mml-book.github.io
-
eferredto by different names
Gee it sure would be easier if we could all agree on a single term
-
three possible outcomes:
There are 4 possible results with 3 different outcomes, so the random variable only depends on the outcomes and not the minutia of the results (as in, it doesn't matter if a result is TH or HT, the outcome is the same)
-
event space
So the outcome {HH, HH, HH} is an event space A of the sample space {HH, TT, HT, TH} and has an associated probability P(A) with it, if I understand the terms correctly.
-
- Sep 2023
-
www.deeplearningbook.org www.deeplearningbook.org
-
they cannot learn theXOR function, wheref([0,1], w) = 1 andf([1,0], w) = 1 butf([1,1], w) = 0andf([0,0], w) = 0
I find the initial criticism to the MLP regrettable; though it's true that single layer MLPs are unable to learn a simple XOR function, this is solved by additional layers. It's unfortunate that this misunderstanding resulted in a dip in popularity in their early days.
-