public void ForwardTest_BiDirectional_RowMajor()
{
const int batchSize = 2;
const int inputSize = 3;
const int numberOfNeurons = 2;
SRNCell layer = new SRNCell(new Shape(new int[] { batchSize, inputSize }), RNNDirection.BiDirectional, numberOfNeurons, MatrixLayout.RowMajor, null);
layer.W.Set(new float[] { 0.57935405f, -0.2018174f, 0.3719957f, -0.11352646f, 0.23978919f, 0.30809408f }); // 3x2 matrix
layer.U.Set(new float[] { -0.6668052f, 0.0096491f }); // 2x1 matrix
layer.B.Set(new float[] { -0.2414839f, -0.08907348f }); // 2x1 vector
Tensor x = new Tensor(null, new[] { batchSize, inputSize });
x.Set(new float[] { -0.1f, 0.2f, 0.3f, 0.4f, -0.5f, 0.6f });
// set expectations
Tensor expected = new Tensor(null, new[] { numberOfNeurons, numberOfNeurons });
expected.Set(new float[] { 0, 0.06266524f, 0.3143638f, 0 });
// calculate
Session session = new Session();
Tensor y = layer.Forward(session, new[] { x })[0];
Helpers.AreTensorsEqual(expected, y);
y.SetGradient(new float[] { 0.1f, 0.2f, 0.3f, 0.4f });
session.Unroll();
Helpers.AreArraysEqual(
new float[] { 0.120000005f, -0.15f, 0.18f, -0.0200000014f, 0.0400000028f, 0.0600000024f, },
layer.W.Gradient);
Helpers.AreArraysEqual(
new float[] { 0f, 0f, },
layer.U.Gradient);
Helpers.AreArraysEqual(
new float[] { 0.3f, 0.2f, },
layer.B.Gradient);
Helpers.AreArraysEqual(
new float[] { -0.0227052923f, 0.0479578376f, 0.0616188161f, 0.17380622f, -0.06054522f, 0.111598708f, },
x.Gradient);
}
public void ForwardTest_BiDirectional_ColumnMajor()
{
const int batchSize = 2;
const int inputSize = 3;
const int numberOfNeurons = 2;
SRNCell layer = new SRNCell(new Shape(new int[] { batchSize, inputSize }), RNNDirection.BiDirectional, numberOfNeurons, MatrixLayout.ColumnMajor, null);
layer.W.Set(new float[] { 0.57935405f, -0.2018174f, 0.3719957f, -0.11352646f, 0.23978919f, 0.30809408f }); // 3x2 matrix
layer.U.Set(new float[] { -0.6668052f, 0.0096491f }); // 1x2 matrix
layer.B.Set(new float[] { -0.2414839f, -0.08907348f }); // 2x1 vector
Tensor x = new Tensor(null, new[] { batchSize, inputSize });
x.Set(new float[] { -0.1f, 0.2f, 0.3f, 0.4f, -0.5f, 0.6f });
// set expectations
Tensor expected = new Tensor(null, new[] { numberOfNeurons, numberOfNeurons });
expected.Set(new float[] { 0, 0.001524193f, 0, 0.07181925f });
// calculate
Session session = new Session();
Tensor y = layer.Forward(session, new[] { x })[0];
Helpers.AreTensorsEqual(expected, y);
y.SetGradient(new float[] { 0.1f, 0.2f, 0.3f, 0.4f });
session.Unroll();
Helpers.AreArraysEqual(
new float[] { 0, 0.14077194f, 0, -0.1609649f, 0, 0.3011579f, },
layer.W.Gradient);
Helpers.AreArraysEqual(
new float[] { 0, 0.01436385f, },
layer.U.Gradient);
Helpers.AreArraysEqual(
new float[] { 0, 0.601929843f, },
layer.B.Gradient);
Helpers.AreArraysEqual(
new float[] { -0.04036348f, -0.0227052923f, 0.0616188161f, -0.08111643f, -0.0456296727f, 0.1238322f, },
x.Gradient);
}
public void ForwardTest_Forward_ColumnMajor()
{
const int batchSize = 2;
const int inputSize = 3;
const int numberOfNeurons = 2;
SRNCell layer = new SRNCell(new Shape(new int[] { batchSize, inputSize }), RNNDirection.ForwardOnly, numberOfNeurons, MatrixLayout.ColumnMajor, null);
layer.W.Set(new float[] { 0.57935405f, -0.2018174f, 0.3719957f, -0.11352646f, 0.23978919f, 0.30809408f }); // 3x2 matrix
layer.U.Set(new float[] { -0.6668052f, 0.0096491f, 0.17214662f, -0.4206545f }); // 2x2 matrix
layer.B.Set(new float[] { -0.2414839f, -0.08907348f }); // 2x1 vector
Tensor x = new Tensor(null, new[] { batchSize, inputSize });
x.Set(new float[] { -0.1f, 0.2f, 0.3f, 0.4f, -0.5f, 0.6f });
// set expectations
Tensor expected = new Tensor(null, new[] { numberOfNeurons, numberOfNeurons });
expected.Set(new float[] { 0f, 0.0008311934f, 0f, 0.07146961f, });
// calculate
Session session = new Session();
Tensor y = layer.Forward(session, new[] { x })[0];
Helpers.AreTensorsEqual(expected, y);
y.SetGradient(new float[] { 0.1f, 0.2f, 0.3f, 0.4f });
session.Unroll();
Helpers.AreArraysEqual(
new float[] { 0f, 0.1568262f, 0f, -0.1936524f, 0f, 0.2495215f, },
layer.W.Gradient);
Helpers.AreArraysEqual(
new float[] { 0f, 0f, 0f, 0.0003324774f, },
layer.U.Gradient);
Helpers.AreArraysEqual(
new float[] { 0f, 0.4317382f, },
layer.B.Gradient);
Helpers.AreArraysEqual(
new float[] { -0.006405319f, -0.003603125f, 0.009778349f, -0.08072696f, -0.04541058f, 0.1232376f, },
x.Gradient);
}
public void ForwardTest_Forward_RowMajor()
{
const int batchSize = 2;
const int inputSize = 3;
const int numberOfNeurons = 2;
SRNCell layer = new SRNCell(new Shape(new int[] { batchSize, inputSize }), RNNDirection.ForwardOnly, numberOfNeurons, MatrixLayout.RowMajor, null);
layer.W.Set(new float[] { 0.57935405f, -0.2018174f, 0.3719957f, -0.11352646f, 0.23978919f, 0.30809408f }); // 3x2 matrix
layer.U.Set(new float[] { -0.6668052f, 0.0096491f, 0.17214662f, -0.4206545f }); // 2x2 matrix
layer.B.Set(new float[] { -0.2414839f, -0.08907348f }); // 2x1 vector
Tensor x = new Tensor(null, new[] { batchSize, inputSize });
x.Set(new float[] { -0.1f, 0.2f, 0.3f, 0.4f, -0.5f, 0.6f });
// set expectations
Tensor expected = new Tensor(null, new[] { numberOfNeurons, numberOfNeurons });
expected.Set(new float[] { 0f, 0.06266524f, 0.3149685f, 0f, });
// calculate
Session session = new Session();
Tensor y = layer.Forward(session, new[] { x })[0];
Helpers.AreTensorsEqual(expected, y);
y.SetGradient(new float[] { 0.1f, 0.2f, 0.3f, 0.4f });
session.Unroll();
Helpers.AreArraysEqual(
new float[] { 0.12f, -0.15f, 0.18f, -0.02028947f, 0.04057895f, 0.06086842f, },
layer.W.Gradient);
Helpers.AreArraysEqual(
new float[] { 0f, 0.01879957f, 0f, 0f, },
layer.U.Gradient);
Helpers.AreArraysEqual(
new float[] { 0.3f, 0.2028947f, },
layer.B.Gradient);
Helpers.AreArraysEqual(
new float[] { -0.02303392f, 0.04865196f, 0.06251067f, 0.1738062f, -0.06054522f, 0.1115987f, },
x.Gradient);
}
public void ForwardBackwardTest1()
{
const int T = 2;
const int N = 3;
Session session = new Session();
SRNCell layer = new SRNCell(new Shape(new[] { -1, N }), RNNDirection.ForwardOnly, 2, MatrixLayout.RowMajor, null);
layer.W.Randomize(this.random);
layer.U.Randomize(this.random);
layer.B.Randomize(this.random);
////layer.W.Set(new float[] { 0.1f, 0.2f, -0.3f, 0.4f, 0.5f, 0.6f }); // 3x2 matrix
////layer.U.Set(new float[] { 0.1f, 0.2f, 0.3f, 0.4f }); // 2x2 matrix
////layer.B.Set(new float[] { 0.1f, 0.2f }); // 2x1 vector
Tensor x = new Tensor(null, new[] { T, N });
x.Randomize(this.random);
////x.Set(new float[] { 0.1f, 0.2f, 0.3f, 0.4f, 0.5f, 0.6f });
IList <Tensor> xs = new[] { x };
IList <Tensor> ys = layer.Forward(session, xs);
float[] bw = layer.B.Weights;
Tensor expected = new Tensor(null, new[] { 2, 2 });
expected.Weights[0] = Matrix.DotProduct(3, layer.W.Weights, 0, x.Weights, 0) + bw[0];
expected.Weights[1] = Matrix.DotProduct(3, layer.W.Weights, 3, x.Weights, 0) + bw[1];
Nonlinearity.ReLU(2, expected.Weights, 0, expected.Weights, 0);
expected.Weights[2] = Matrix.DotProduct(3, layer.W.Weights, 0, x.Weights, 3) + bw[0] + Matrix.DotProduct(2, layer.U.Weights, 0, expected.Weights, 0);
expected.Weights[3] = Matrix.DotProduct(3, layer.W.Weights, 3, x.Weights, 3) + bw[1] + Matrix.DotProduct(2, layer.U.Weights, 2, expected.Weights, 0);
Nonlinearity.ReLU(2, expected.Weights, 2, expected.Weights, 2);
Helpers.AreTensorsEqual(expected, ys[0]);
// unroll the graph
////session.GetGradient(ys[0]).Randomize(this.random);
ys[0].SetGradient(new float[] { 0.1f, 0.2f, 0.3f, 0.4f });
session.Unroll();
////float[] dy = session.GetGradient(ys[0]).Weights.ToArray();
float[] dy = new float[] { 0.1f, 0.2f, 0.3f, 0.4f };
float[] expectedWG = new float[layer.W.Length];
float[] expectedUG = new float[layer.U.Length];
float[] expectedBG = new float[layer.B.Length];
float[] expectedDx = new float[x.Length];
for (int oi = 2, ii = 3; oi >= 0; oi -= 2, ii -= 3)
{
Nonlinearity.ReLUGradient(2, dy, oi, true, expected.Weights, oi, dy, oi);
// should be x' * dy
Matrix.VxV(MatrixLayout.ColumnMajor, 3, 2, x.Weights, ii, dy, oi, expectedWG, 0, false);
// should be W' * dy
Matrix.MxV(MatrixLayout.ColumnMajor, 3, 2, layer.W.Weights, 0, false, dy, oi, expectedDx, ii, true);
if (oi > 0)
{
// should be x(t-1)' * dy
Matrix.VxV(MatrixLayout.ColumnMajor, 2, 2, expected.Weights, oi - 2, dy, oi, expectedUG, 0, false);
// should be U' * dy
Matrix.MxV(MatrixLayout.ColumnMajor, 2, 2, layer.U.Weights, 0, false, dy, oi, dy, oi - 2, false);
////MKL.MxV(MatrixLayout.RowMajor, 2, 2, layer.U.Weights, 0, false, dy, oi, dy, oi - 2, false);
}
// should be dy
Mathematics.Add(2, dy, oi, expectedBG, 0);
}
Helpers.AreArraysEqual(expectedWG, layer.W.Gradient);
////Helpers.AreArraysEqual(expectedUG, layer.U.Gradient);
Helpers.AreArraysEqual(expectedBG, layer.B.Gradient);
Helpers.AreArraysEqual(expectedDx, x.Gradient);
}