@alfcnz Amazing, how did you make this?
-
-
- Još 4 druga odgovora
Novi razgovor -
-
-
What a treasure! Thought it’s a little unnatural for me to imagine a 5D situation
one more, as mentioned in other comments, could I say some multiplication and addition are equal to stretching space? -
Some multiplications and additions only rotate it. You need a non linearity in order to have a localised (cf. global) transformation.
- Još 2 druga odgovora
Novi razgovor -
-
-
But the stretching may be a complicated computation. It simply learns a certain kind of very general computation.
-
Not very complicated, a few multiplications, a few additions, and some zeroing of negative numbers.
- Još 2 druga odgovora
Novi razgovor -
-
-
A silly question. You represent the stretched space, but not the data. Is the reason for this the data being a point set?
-
(Question less in why the processing being made, more in the representation itself..)
- Još 3 druga odgovora
Novi razgovor -
-
-
Very nice and insightful! One question would be: are 100 neurons needed? Can it be done with 10?
-
More about this next class! The answer is: it depends. For a 3 layer net yes, 100 neurons are indeed needed. Anyhow, you can play with my notebook:https://github.com/Atcold/pytorch-Deep-Learning-Minicourse/blob/master/04-spiral_classification.ipynb …
Kraj razgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.
, and assistant professor of Computer Science at New York University