public Tensor categorical_crossentropy(Tensor target, Tensor output, bool from_logits = false)
{
// https://github.com/fchollet/keras/blob/f65a56fb65062c8d14d215c9f4b1015b97cc5bf3/keras/backend/cntk_backend.py#L1480
var _output = In(output);
var _target = In(target);
if (from_logits)
{
var result = C.CrossEntropyWithSoftmax(_output, _target);
// cntk's result shape is (batch, 1), while keras expect (batch, )
CNTK.Function r = C.Reshape(result, NDShape.CreateNDShape(new int[] { }));
return(Out(r));
}
else
{
// scale preds so that the class probas of each sample sum to 1
var o = C.ElementDivide(_output.function, C.ReduceSum(_output, Axis.EndStaticAxis()));
var eps = Constant.Scalar(epsilon(), DeviceDescriptor.CPUDevice);
var omeps = Constant.Scalar(1.0 - epsilon(), DeviceDescriptor.CPUDevice);
// avoid numerical instability with _EPSILON clipping
o = C.Clip(o, eps, omeps);
CNTK.Function r = C.Negate(C.ReduceSum(C.ElementTimes(_target, C.Log(_output)), Axis.EndStaticAxis()));
return(Out(r));
}
}
C.Function squash(C.Function vectors, string name, int axis)
{
var squared_values = CC.Square(vectors);
var s_squared_sum = CC.ReduceSum(squared_values, new C.AxisVector(new C.Axis[] { new C.Axis(axis) }), keepDims: true);
var epsilon = C.Constant.Scalar(C.DataType.Float, 1e-7, computeDevice);
var one = C.Constant.Scalar(C.DataType.Float, 1.0, computeDevice);
var normalize_factor = CC.Plus(CC.Sqrt(s_squared_sum), epsilon);
var one_plus_s_squared_sum = CC.Plus(s_squared_sum, one);
var scale = CC.ElementDivide(s_squared_sum, one_plus_s_squared_sum);
scale = CC.ElementDivide(scale, normalize_factor);
var result = CC.ElementTimes(scale, vectors, name);
return(result);
}
public Tensor Div(float a, Tensor b)
{
return(Out(C.ElementDivide(In(a, b.Shape), In(b))));
}
public Tensor Div(Tensor a, float b)
{
return(Out(C.ElementDivide(In(a), In(b, a.Shape))));
}
public Tensor Div(Tensor a, Tensor b)
{
return(Out(C.ElementDivide(In(a), In(b))));
}
public Tensor div <T>(Tensor a, T b)
{
return(Out(C.ElementDivide(In(a), InGeneric(b))));
}
void create_network()
{
imageVariable = Util.inputVariable(input_shape, "image");
var conv1 = Layers.Convolution2D(
imageVariable, 256, new int[] { 9, 9 }, computeDevice,
use_padding: false, activation: CC.ReLU, name: "conv1");
var primarycaps = create_primary_cap(
conv1, dim_capsule: 8, n_channels: 32,
kernel_size: new int[] { 9, 9 }, strides: new int[] { 2, 2 }, pad: false);
var digitcaps = create_capsule_layer(
primarycaps, num_capsule: 10, dim_capsule: 16,
routings: routings, name: "digitcaps");
var out_caps = get_length_and_remove_last_dimension(digitcaps, name: "capsnet");
categoricalLabel = Util.inputVariable(new int[] { 10 }, "label");
var masked_by_y = get_mask_and_infer_from_last_dimension(digitcaps, CC.Combine(new C.VariableVector()
{
categoricalLabel
}));
var masked = get_mask_and_infer_from_last_dimension(digitcaps, null);
var decoder = create_decoder(masked.Output.Shape.Dimensions.ToArray());
var decoder_output_training = Model.invoke_model(decoder, new C.Variable[] { masked_by_y });
var decoder_output_evaluation = Model.invoke_model(decoder, new C.Variable[] { masked });
network = CC.Combine(new C.VariableVector()
{
out_caps, decoder_output_training
}, "overall_training_network");
Logging.log_number_of_parameters(network);
// first component of the loss
var y_true = categoricalLabel;
var y_pred = out_caps;
var digit_loss = CC.Plus(
CC.ElementTimes(y_true, CC.Square(CC.ElementMax(DC(0), CC.Minus(DC(0.9), y_pred), ""))),
CC.ElementTimes(DC(0.5),
CC.ElementTimes(CC.Minus(DC(1), y_true), CC.Square(CC.ElementMax(DC(0), CC.Minus(y_pred, DC(0.1)), "")))));
digit_loss = CC.ReduceSum(digit_loss, C.Axis.AllStaticAxes());
// second component of the loss
var num_pixels_at_output = Util.np_prod(decoder_output_training.Output.Shape.Dimensions.ToArray());
var squared_error = CC.SquaredError(decoder_output_training, imageVariable);
var image_mse = CC.ElementDivide(squared_error, DC(num_pixels_at_output));
loss_function = CC.Plus(digit_loss, CC.ElementTimes(DC(0.35), image_mse));
eval_function = CC.ClassificationError(y_pred, y_true);
learner = CC.AdamLearner(
new C.ParameterVector(network.Parameters().ToArray()),
new C.TrainingParameterScheduleDouble(0.001 * batch_size, (uint)batch_size),
new C.TrainingParameterScheduleDouble(0.9),
true,
new C.TrainingParameterScheduleDouble(0.99));
trainer = CC.CreateTrainer(network, loss_function, eval_function, new C.LearnerVector(new C.Learner[] { learner }));
evaluator = CC.CreateEvaluator(eval_function);
}