def test_squareblock2D(self):
data=np.array([[
[[1,2,3,4],
[5,6,7,8],
[9,10,11,12],
[13,14,15,16]]]],'f')
out=learn.empty((1,1,2,2))
learn.squareblock2D(data,out,2)
self.assertTrue(np.allclose(out,np.array([[[[8.1240387,11.74734211],[23.36664391,27.31299973]]]],'f')))
def test_sqrlayer(self):
data=np.array([
[[[1,-1]],
[[0,0]]],
[[[1,-1]],
[[2,-2]]]],'f')
out=learn.empty((2,1,1,2))
learn.squarelayer(data,out,2)
self.assertTrue(np.allclose(out,np.array([[[[1,1]]],[[[2.23607039,2.23607039]]]],'f')))
def test_maxblock2D(self):
data=np.array([[
[[1,2,3,4],
[5,6,7,8],
[9,10,11,12],
[13,14,15,16]]]],'f')
out=learn.empty((1,1,2,2))
learn.maxblock2D(data,out,2)
self.assertTrue(np.array_equal(out,np.array([[[[6,8],[14,16]]]],'f')))
def test_linearpoolin(self):
data=np.array([[
[[1,2,3,4],
[5,6,7,8],
[9,10,11,12],
[13,14,15,16]]]],'f')
out=learn.empty((1,1,2,2))
learn.linearpoolin(data,out)
self.assertTrue(np.array_equal(out,np.array([[[[54,63],[90,99]]]],'f')))
def test_forwardmax2D(self):
data=np.array([[
[[1,2,3,4],
[5,6,7,8],
[9,10,11,12],
[13,14,15,16]]]],'f')
grad=-data
out=learn.empty((1,1,2,2))
learn.followmaxblock2D(grad,data,out,2)
self.assertTrue(np.array_equal(out,np.array([[[[-6,-8],[-14,-16]]]],'f')))
def test_revmax2D(self):
data=np.array([[
[[1,2,3,4],
[5,6,7,8],
[9,10,11,12],
[13,14,15,16]]]],'f')
grad=np.array([[[[1,2],[3,4]]]],'f')
out=learn.empty((1,1,4,4))
learn.reversemaxblock2D(grad,data,out,2)
self.assertTrue(np.array_equal(out,np.array([[[[0,0,0,0],[0,1,0,2],[0,0,0,0],[0,3,0,4]]]],'f')))
def test_revsqrlayer(self):
grad=np.array([[[[1,2]]],[[[3,4]]]],'f')
data=np.array([
[[[1,-1]],
[[0,0]]],
[[[1,-1]],
[[2,-2]]]],'f')
outdata=np.array([[[[1,1]]],[[[2.23607039,2.23607039]]]],'f')
out=learn.empty((2,2,1,2))
learn.reversesquarelayer(grad,data,outdata,out,2)
self.assertTrue(np.allclose(out,np.array([[[[1,-2]],[[0,0]]],[[[1.34163928,-1.78885245]],[[2.68327856,-3.57770491]]]],'f')))
def test_linearpoolout(self):
data=np.array([[
[[1,2],
[3,4]]]],'f')
out=learn.empty((1,1,4,4))
learn.linearpoolout(data,out)
self.assertTrue(np.array_equal(out,np.array([[[
[1,3,3,2],
[4,10,10,6],
[4,10,10,6],
[3,7,7,4]]]],'f')))
def test_forwardsqr2D(self):
data=np.array([
[[[1,-1]],
[[0,0]]],
[[[1,-1]],
[[2,-2]]]],'f')
outdata=np.array([[[[1,1]]],[[[2.23607039,2.23607039]]]],'f')
grad=np.ones((2,2,1,2),'f')
out=learn.empty((2,1,1,2))
learn.followsquarelayer(grad,data,outdata,out,2)
self.assertTrue(np.allclose(out,np.array([[[[1,-1]]],[[[1.34163928,-1.34163928]]]],'f')))
def test_forwardsqr2D(self):
data=np.array([[
[[1,2,3,4],
[5,6,7,8],
[9,10,11,12],
[13,14,15,16]]]],'f')
outd=np.array([[[[8.1240387,11.74734211],[23.36664391,27.31299973]]]],'f')
grad=np.ones((1,1,4,4),'f')
out=learn.empty((1,1,2,2))
learn.followsquareblock2D(grad,data,outd,out,2)
self.assertTrue(np.allclose(out,np.array([[[[1.72328091,1.87276411],[1.96861827,1.97708058]]]],'f')))
def test_gradconv4D(self):
data=np.array([
[[[1,-1],[2,-2],[3,-3]],[[4,-4],[5,-5],[6,-6]],[[7,-7],[8,-8],[9,-9]]],
[[[10,-10],[20,-20],[30,-30]],[[40,-40],[50,-50],[60,-60]],[[70,-70],[80,-80],[90,-90]]]
],'f')
grad=np.array([
[[[1,-1],[-1,1]],[[-2,-3],[3,2]],[[-1,3],[-2,1]],[[-1,4],[-4,4]]],
[[[1,-1],[-1,1]],[[-2,-3],[3,2]],[[-1,3],[-2,1]],[[-1,4],[-4,4]]]
],'f')
out=learn.empty((4,2,2,2))
learn.gradconvolution4D(data,grad,out)
self.assertTrue(np.array_equal(out,np.array([
[[[0,0],[0,0]],[[0,0],[0,0]]],
[[[154,-154],[154,-154]],[[154,-154],[154,-154]]],
[[[22,-22],[33,-33]],[[55,-55],[66,-66]]],
[[[121,-121],[154,-154]],[[220,-220],[253,-253]]]],'f')))
def test_gradnoconv(self):
data=np.array([
[[[1,-1],[2,-2],[3,-3]],
[[4,-4],[5,-5],[6,-6]],
[[7,-7],[8,-8],[9,-9]]],
[[[10,-10],[20,-20],[30,-30]],
[[40,-40],[50,-50],[60,-60]],
[[70,-70],[80,-80],[90,-90]]]
],'f')
grad=np.array([[[[1]]],[[[1]]]],'f')
filters=learn.empty((1,1,1,3,3,2))
learn.grad_noconv4D(data,grad,filters)
self.assertTrue(np.array_equal(filters,np.array(
[[[[[[11,-11],[22,-22],[33,-33]],
[[44,-44],[55,-55],[66,-66]],
[[77,-77],[88,-88],[99,-99]]]]]],'f')))
def test_revsqr2D(self):
data=np.array([[
[[1,2,3,4],
[5,6,7,8],
[9,10,11,12],
[13,14,15,16]]]],'f')
outd=np.array([[[[8.1240387,11.74734211],[23.36664391,27.31299973]]]],'f')
grad=np.array([[[[1,1],[1,1]]]],'f')
out=learn.empty((1,1,4,4))
learn.reversesquareblock2D(grad,data,outd,out,2)
self.assertTrue(np.allclose(out,np.array(
[[[[ 0.12309149, 0.24618298, 0.25537694, 0.34050256],
[ 0.61545742, 0.73854893, 0.5958795, 0.68100512],
[ 0.38516444, 0.42796049, 0.40273863, 0.43935126],
[ 0.55634862, 0.5991447, 0.54918903, 0.58580166]]]]
,'f')))
def test_fromhidden(self):
hidden=np.array([[[[1]]],[[[10]]]],'f')
filters=np.array(
[[[
[[[1,1],[1,0],[1,-1]],
[[0,1],[0,0],[0,-1]],
[[-1,1],[-1,0],[-1,-1]]]
]]],'f')
data=learn.empty((2,3,3,2))
learn.fromhidden_noconv4D(hidden,filters,data)
self.assertTrue(np.array_equal(data,np.array([
[[[1,1],[1,0],[1,-1]],
[[0,1],[0,0],[0,-1]],
[[-1,1],[-1,0],[-1,-1]]],
[[[10,10],[10,0],[10,-10]],
[[0,10],[0,0],[0,-10]],
[[-10,10],[-10,0],[-10,-10]]]],'f')))
def test_tohidden(self):
data=np.array([
[[[1,-1],[2,-2],[3,-3]],
[[4,-4],[5,-5],[6,-6]],
[[7,-7],[8,-8],[9,-9]]],
[[[10,-10],[20,-20],[30,-30]],
[[40,-40],[50,-50],[60,-60]],
[[70,-70],[80,-80],[90,-90]]]
],'f')
filters=np.array(
[[[
[[[1,1],[1,0],[1,-1]],
[[0,1],[0,0],[0,-1]],
[[-1,1],[-1,0],[-1,-1]]]
]]],'f')
out=learn.empty((2,1,1,1))
learn.tohidden_noconv4D(data,filters,out)
self.assertTrue(np.array_equal(out,np.array([[[[-12]]],[[[-120]]]],'f')))
def test_revconv4D(self):
filters=np.array([
[[[1,-1],[1,-1]],[[1,-1],[1,-1]]],
[[[-1,1],[-1,1]],[[-1,1],[-1,1]]],
[[[1,-1],[1,-1]],[[-1,1],[-1,1]]],
[[[1,-1],[-1,1]],[[1,-1],[-1,1]]]],'f')
grad=np.array([
[[[1,-1],[-1,1]],[[-2,-3],[3,2]],[[-1,3],[-2,1]],[[-1,4],[-4,4]]],
[[[0,-1],[-1,1]],[[-2,-3],[3,2]],[[-1,3],[-2,1]],[[-1,4],[-4,4]]]
],'f')
data=learn.empty((2,3,3,2))
learn.reverseconvolution4D(grad,filters,data)
self.assertTrue(np.array_equal(data,np.array([
[[[1,-1],[12,-12],[1,-1]],
[[-7,7],[10,-10],[-9,9]],
[[-6,6],[4,-4],[-6,6]]],
[[[0,0],[11,-11],[1,-1]],
[[-8,8],[9,-9],[-9,9]],
[[-6,6],[4,-4],[-6,6]]]])))
def test_conv4D(self):
data=np.array([
[[[1,-1],[2,-2],[3,-3]],
[[4,-4],[5,-5],[6,-6]],
[[7,-7],[8,-8],[9,-9]]],
[[[10,-10],[20,-20],[30,-30]],
[[40,-40],[50,-50],[60,-60]],
[[70,-70],[80,-80],[90,-90]]]
],'f')
filters=np.array([
[[[1,-1],[1,-1]],[[1,-1],[1,-1]]],
[[[-1,1],[-1,1]],[[-1,1],[-1,1]]],
[[[1,-1],[1,-1]],[[-1,1],[-1,1]]],
[[[1,-1],[-1,1]],[[1,-1],[-1,1]]]],'f')
out=learn.empty((2,4,2,2))
learn.convolution4D(data,filters,out)
self.assertTrue(np.array_equal(out,np.array([
[[[24,32],[48,56]],[[-24,-32],[-48,-56]],[[-12,-12],[-12,-12]],[[-4,-4],[-4,-4]]],
[[[240,320],[480,560]],[[-240,-320],[-480,-560]],[[-120,-120],[-120,-120]],[[-40,-40],[-40,-40]]]
],'f')))
def test_transform(self):
a0=np.array([-1000,1000],'f')
o=learn.empty((2,))
learn.transform('a sigmoid',o,a0)
self.assertTrue(np.allclose(o,np.array([0,1],'f')))