I'm trying to slice a 2D array using coordinates given by another array.
Example :
def example(p, Q):
a = T.tensor4('a')
q = T.matrix('q')
b = q[:, 0]
c = q[:, 1]
b_i = T.cast(b, 'int32')
c_i = T.cast(c, 'int32')
d, _ = theano.map(lambda pix, x, y : pix[0, x-1:x+2, y-1:y+2],
sequences=[a, b_i, c_i])
f = theano.function([a, q], d, allow_input_downcast=True)
return f(p, Q)
# test
p = np.arange(200).reshape((2, 1, 10, 10))
Q = np.array([[3, 4], [5, 6]])
print p
answer = example(p, Q)
print answer
=========================================
Gives :
[[[[ 0 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]
[190 191 192 193 194 195 196 197 198 199]]]]
[[[ 23. 24. 25.]
[ 33. 34. 35.]
[ 43. 44. 45.]]
[[ 145. 146. 147.]
[ 155. 156. 157.]
[ 165. 166. 167.]]]
Which is the correct output.
However :
p = np.arange(200).reshape((2, 1, 10, 10))
Q = np.array([[3, 4], [8, 9]])
print p
answer = example5(p, Q)
print answer
crashes because the coordinate [8, 9] is near the edge of the 10x10 array
whereas [3, 4] isn't. So the two iterations in scan give outputs of two
different shapes. So it crashes.
How can I solve this problem (either by zero-padding somewhere, or wrapping
around the slice) ? I'm not too bothered about exactness, I just need a
solution that doesn't crash.
--
---
You received this message because you are subscribed to the Google Groups
"theano-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
