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
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
|
local Euclidean, parent = torch.class('nn.Euclidean', 'nn.Module')
function Euclidean:__init(inputSize,outputSize)
parent.__init(self)
self.weight = torch.Tensor(inputSize,outputSize)
self.gradWeight = torch.Tensor(inputSize,outputSize)
-- state
self.gradInput:resize(inputSize)
self.output:resize(outputSize)
self.fastBackward = true
self:reset()
end
function Euclidean:reset(stdv)
if stdv then
stdv = stdv * math.sqrt(3)
else
stdv = 1./math.sqrt(self.weight:size(1))
end
if nn.oldSeed then
for i=1,self.weight:size(2) do
self.weight:select(2, i):apply(function()
return torch.uniform(-stdv, stdv)
end)
end
else
self.weight:uniform(-stdv, stdv)
end
end
local function view(res, src, ...)
local args = {...}
if src:isContiguous() then
res:view(src, table.unpack(args))
else
res:reshape(src, table.unpack(args))
end
end
function Euclidean:updateOutput(input)
-- lazy initialize buffers
self._input = self._input or input.new()
self._weight = self._weight or self.weight.new()
self._expand = self._expand or self.output.new()
self._expand2 = self._expand2 or self.output.new()
self._repeat = self._repeat or self.output.new()
self._repeat2 = self._repeat2 or self.output.new()
local inputSize, outputSize = self.weight:size(1), self.weight:size(2)
-- y_j = || w_j - x || = || x - w_j ||
if input:dim() == 1 then
view(self._input, input, inputSize, 1)
self._expand:expandAs(self._input, self.weight)
self._repeat:resizeAs(self._expand):copy(self._expand)
self._repeat:add(-1, self.weight)
self.output:norm(self._repeat, 2, 1)
self.output:resize(outputSize)
elseif input:dim() == 2 then
local batchSize = input:size(1)
view(self._input, input, batchSize, inputSize, 1)
self._expand:expand(self._input, batchSize, inputSize, outputSize)
-- make the expanded tensor contiguous (requires lots of memory)
self._repeat:resizeAs(self._expand):copy(self._expand)
self._weight:view(self.weight, 1, inputSize, outputSize)
self._expand2:expandAs(self._weight, self._repeat)
if torch.type(input) == 'torch.CudaTensor' then
-- requires lots of memory, but minimizes cudaMallocs and loops
self._repeat2:resizeAs(self._expand2):copy(self._expand2)
self._repeat:add(-1, self._repeat2)
else
self._repeat:add(-1, self._expand2)
end
self.output:norm(self._repeat, 2, 2)
self.output:resize(batchSize, outputSize)
else
error"1D or 2D input expected"
end
return self.output
end
function Euclidean:updateGradInput(input, gradOutput)
if not self.gradInput then
return
end
self._div = self._div or input.new()
self._output = self._output or self.output.new()
self._gradOutput = self._gradOutput or input.new()
self._expand3 = self._expand3 or input.new()
if not self.fastBackward then
self:updateOutput(input)
end
local inputSize, outputSize = self.weight:size(1), self.weight:size(2)
--[[
dy_j -2 * (w_j - x) x - w_j
---- = --------------- = -------
dx 2 || w_j - x || y_j
--]]
-- to prevent div by zero (NaN) bugs
self._output:resizeAs(self.output):copy(self.output):add(0.0000001)
view(self._gradOutput, gradOutput, gradOutput:size())
self._div:cdiv(gradOutput, self._output)
if input:dim() == 1 then
self._div:resize(1, outputSize)
self._expand3:expandAs(self._div, self.weight)
if torch.type(input) == 'torch.CudaTensor' then
self._repeat2:resizeAs(self._expand3):copy(self._expand3)
self._repeat2:cmul(self._repeat)
else
self._repeat2:cmul(self._repeat, self._expand3)
end
self.gradInput:sum(self._repeat2, 2)
self.gradInput:resizeAs(input)
elseif input:dim() == 2 then
local batchSize = input:size(1)
self._div:resize(batchSize, 1, outputSize)
self._expand3:expand(self._div, batchSize, inputSize, outputSize)
if torch.type(input) == 'torch.CudaTensor' then
self._repeat2:resizeAs(self._expand3):copy(self._expand3)
self._repeat2:cmul(self._repeat)
else
self._repeat2:cmul(self._repeat, self._expand3)
end
self.gradInput:sum(self._repeat2, 3)
self.gradInput:resizeAs(input)
else
error"1D or 2D input expected"
end
return self.gradInput
end
function Euclidean:accGradParameters(input, gradOutput, scale)
local inputSize, outputSize = self.weight:size(1), self.weight:size(2)
scale = scale or 1
--[[
dy_j 2 * (w_j - x) w_j - x
---- = --------------- = -------
dw_j 2 || w_j - x || y_j
--]]
-- assumes a preceding call to updateGradInput
if input:dim() == 1 then
self.gradWeight:add(-scale, self._repeat2)
elseif input:dim() == 2 then
self._sum = self._sum or input.new()
self._sum:sum(self._repeat2, 1)
self._sum:resize(inputSize, outputSize)
self.gradWeight:add(-scale, self._sum)
else
error"1D or 2D input expected"
end
end
function Euclidean:type(type, tensorCache)
if type then
-- prevent premature memory allocations
self:clearState()
end
return parent.type(self, type, tensorCache)
end
function Euclidean:clearState()
nn.utils.clear(self, {
'_input',
'_output',
'_gradOutput',
'_weight',
'_div',
'_sum',
'_expand',
'_expand2',
'_expand3',
'_repeat',
'_repeat2',
})
return parent.clearState(self)
end
|
