/
dropout.cr
54 lines (52 loc) · 1.92 KB
/
dropout.cr
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
# Copyright (c) 2021 Crystal Data Contributors
#
# MIT License
#
# Permission is hereby granted, free of charge, to any person obtaining
# a copy of this software and associated documentation files (the
# "Software"), to deal in the Software without restriction, including
# without limitation the rights to use, copy, modify, merge, publish,
# distribute, sublicense, and/or sell copies of the Software, and to
# permit persons to whom the Software is furnished to do so, subject to
# the following conditions:
#
# The above copyright notice and this permission notice shall be
# included in all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
# NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE
# LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
# OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
# WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
module Num::NN
# Computes a forward dropout activation
#
# ## Arguments
#
# * input : `Tensor` - `Tensor` to activate
# * mask : `Tensor` - Mask to dropout
# * probability : `Float` - Probability of dropout
def dropout(
input : Tensor(U, OCL(U)),
mask : Tensor(U, OCL(U)),
probability : Float
) : Tensor(U, OCL(U)) forall U
input * mask / U.new(probability)
end
# Computes a backwards dropout derivative
#
# ## Arguments
#
# * gradient : `Tensor` - `Tensor` used to compute backwards pass
# * mask : `Tensor` - Mask to apply to the gradient
# * probability : `Float` - Probability of dropout
def dropout_backwards(
gradient : Tensor(U, CPU(U)),
mask : Tensor(U, CPU(U)),
probability : Float
) : Tensor(U, OCL(U)) forall U
gradient * mask / U.new(probability)
end
end