If a difference is bigger than this in log terms, then the sum or difference of them will just be the larger (to 12 or so decimal places for double, and 7 or 8 for float).
Puts a softmax layer over a collection of scores, so they look like probabilities
Puts a softmax layer over a collection of scores, so they look like probabilities
A collection of unnormalized scores
Indicates how spiked the probability distribution should be
Puts a softmax layer over a collection of scores, so they look like probabilities
Puts a softmax layer over a collection of scores, so they look like probabilities
A collection of unnormalized scores
Indicates how spiked the probability distribution should be
Returns the log of the portion between fromIndex
, inclusive, and
toIndex
, exclusive, of an array of numbers, which are
themselves input in log form.
Returns the log of the portion between fromIndex
, inclusive, and
toIndex
, exclusive, of an array of numbers, which are
themselves input in log form. This is all natural logarithms.
Reasonable care is taken to do this as efficiently as possible
(under the assumption that the numbers might differ greatly in
magnitude), with high accuracy, and without numerical overflow. Throws an
IllegalArgumentException if logInputs
is of length zero.
Otherwise, returns Double.NegativeInfinity if fromIndex
>=
toIndex
.
Numbers in log form
Start offset (inclusive)
End offset (exclusive)
log(x1 + ... + xn)
sample howMany elements uniformly from xs.
sample howMany elements uniformly from xs. Doesn't retain order of xs
Puts a softmax layer over a collection of scores, so they look like probabilities
Puts a softmax layer over a collection of scores, so they look like probabilities
A collection of unnormalized scores
Indicates how spiked the probability distribution should be
Puts a softmax layer over a collection of scores, so they look like probabilities
Puts a softmax layer over a collection of scores, so they look like probabilities
A collection of unnormalized scores
Indicates how spiked the probability distribution should be
Math utility methods useful for stats and ML User: mihais, dfried Date: 4/23/13