Loading TOC...

cntk functions

The table below lists all the cntk built-in functions (in this namespace: ).

CNTK library integration, providing machine learning features for MarkLogic server.

328 functions
Function name Description
cntk:abs Create an instance of the CNTK built-in elementwise abs operation on the input operands.
cntk:acos Create an instance of the CNTK built-in elementwise acos operation on the input operands.
cntk:ada-delta-learner Constructs an AdaDelta learner.
cntk:ada-grad-learner Constructs an AdaGrad learner.
cntk:adam-learner Constructs an Adam learner.
cntk:alias Create a new Function instance which just aliases the specified ‘x’ Function/Variable such that the ‘Output’ of the new ‘Function’ is same as the ‘Output’ of the specified ‘x’ Function/Variable, and has the newly specified name.
cntk:all-axes Axis object representing all the axes (static and dynamic) of an operand.
cntk:all-static-axes Axis object representing all the static axes of an operand.
cntk:argmax Create an instance of the CNTK built-in argmax operation on specified tensor input operand along the specified axis
cntk:argmin Computes the argmin of the input tensor’s elements across the specified axis.
cntk:as-block Create a new block Function instance which just encapsulates the specified composite Function to create a new Function that appears to be a primitive.
cntk:as-composite Creates a composite Function that has the specified rootFunction as its root.
cntk:asin Create an instance of the CNTK built-in elementwise asin operation on the input operands.
cntk:asinh Create an instance of the CNTK built-in elementwise asinh operation on the input operands.
cntk:assign Assign the value in operand to ref and return the new value, ref need to be the same layout as operand.
cntk:atan Create an instance of the CNTK built-in elementwise atan operation on the input operands.
cntk:atanh Create an instance of the CNTK built-in elementwise atanh operation on the input operands.
cntk:average-pooling-layer Layer factory function to create a average-pooling layer.
cntk:axis Construct an Axis object denoting a static axis with the specified index.
cntk:axis-is-batch Returns true if the axis is of type batch and false otherwise.
cntk:axis-is-dynamic Returns true if the axis is of type dynamic and false otherwise.
cntk:axis-is-ordered Returns True if the axis is ordered; i.e.
cntk:axis-is-sequence Returns true if the axis is of type sequence and false otherwise.
cntk:axis-is-static Returns true if the axis is of type static and false otherwise.
cntk:axis-name Returns the name of this axis.
cntk:axis-static-index Returns True if the axis is of type static and False otherwise
cntk:batch Constructs a batch of cntk:value as a cntk:value.
cntk:batch-normalization Normalizes layer outputs for every minibatch for each output (feature) independently and applies affine transformation to preserve representation of the layer.
cntk:batch-normalization-layer Layer factory function to create a batch-normalization layer.
cntk:batch-of-sequences Constructs a batch of CNTK sequences as a cntk:value.
cntk:bernoulli-random Create an instance of a Bernoulli random operation.
cntk:bernoulli-random-like Create an instance of a Bernoulli random operation.
cntk:bilinear-initializer Returns a bilinear parameter initializer.
cntk:binary-cross-entropy Computes the binary cross entropy (aka logistic loss) between the output and target.
cntk:broadcast-as Creates a sequence out of a non-sequence by endowing the operand with dynamic axes of the same type as the broadcast_as_operand and broadcasting the value of the operand along those dynamic axes.
cntk:ceil Create an instance of the CNTK built-in elementwise ceil operation on the input operands.
cntk:classification-error This operation computes the classification error.
cntk:clip Computes a tensor with all of its values clipped to fall between min_value and max_value, i.e.
cntk:combine Create a new Function instance which just combines the outputs of the specified list of ‘operands’ Functions such that the ‘Outputs’ of the new ‘Function’ are union of the ‘Outputs’ of each of the specified ‘operands’ Functions.
cntk:concatenate-shape Creates a new shape, concatenating the two supplied shapes.
cntk:constant Constructs a constant from a scalar value.
cntk:constant-clone-as Clone the constant as a different data type.
cntk:constant-from-value Constructs a constant from a cntk:value, that is not a sequence or batch.
cntk:constant-initializer Returns a constant parameter initializer.
cntk:constant-record-value-update After setting the value of constant, update the record.
cntk:constant-set-value Sets the value of the constant using a cntk:value that is not a sequence or batch.
cntk:constant-value Returns the value of the constant.
cntk:convolution Computes the convolution of convolution_map (typically a tensor of learnable parameters) with operand (commonly an image or output of a previous convolution/pooling operation).
cntk:convolution-1d-layer Layer factory function to create a convolution-1D layer.
cntk:convolution-2d-layer Layer factory function to create a convolution-2D layer.
cntk:convolution-3d-layer Layer factory function to create a convolution-3D layer.
cntk:convolution-layer Layer factory function to create a convolution layer.
cntk:convolution-transpose Layer factory function to create a convolution transpose layer.
cntk:convolution-transpose-1d-layer Layer factory function to create a convolution-transpose-1D layer.
cntk:convolution-transpose-2d-layer Layer factory function to create a convolution-transpose-2D layer.
cntk:convolution-transpose-3d-layer Layer factory function to create a convolution-transpose-3D layer.
cntk:convolution-transpose-layer Layer factory function to create a convolution-transpose layer.
cntk:convolution-transpose-with-explicit-padding Convolution transpose with explicit lower and upper pad values
cntk:cos Create an instance of the CNTK built-in elementwise cos operation on the input operands.
cntk:cosh Create an instance of the CNTK built-in elementwise cosh operation on the input operands.
cntk:cosine-distance Create an instance of the CNTK built-in operation to compute the cosine distance for the specified input operands.
cntk:cosine-distance-with-negative-samples Given minibatches for x and y, this function computes for each element in x the cosine distance between it and the corresponding y and additionally the cosine distance between x and some other elements of y (referred to a negative samples).
cntk:cpu Returns the CPU device.
cntk:cross-entropy-with-softmax This operation computes the cross entropy between the target_vector and the softmax of the output_vector.
cntk:custom-proxy-op A proxy node that helps saving a model with different number of operands.
cntk:default-batch-axis Returns an Axis object representing the batch axis
cntk:default-device Returns the default device.
cntk:default-input-variable-dynamic-axes Returns the default input variable dynamic axes.
cntk:dense-layer The creates a dense layer and apply it to operand.
cntk:depth-to-space Rearranges elements in the input tensor from the depth dimension into spatial blocks.
cntk:dropout Each element of the input is independently set to 0 with probability dropout_rate or to 1 / (1 - dropout_rate) times its original value (with probability 1-dropout_rate).
cntk:dropout-layer Layer factory function to create a dropout layer.
cntk:dynamic-axis Creates a dynamic axis.
cntk:edit-distance-error Edit distance error evaluation function with the option of specifying penalty of substitution, deletion and insertion, as well as squashing the input sequences and ignoring certain samples.
cntk:element-and Create an instance of the CNTK built-in elementwise AND logical operation on the input operands.
cntk:element-divide Create an instance of the CNTK built-in elementwise division operation on specified tensor input operands.
cntk:element-max Compute the element wise maximum operation between the given operands.
cntk:element-min Compute the element wise minimum operation between the given operands.
cntk:element-not Create an instance of the CNTK built-in elementwise NOT logical operation on the input operands.
cntk:element-or Create an instance of the CNTK built-in elementwise OR logical operation on the input operands.
cntk:element-select Return either value_if_true or value_if_false based on the value of flag.
cntk:element-times Create an instance of the CNTK built-in elementwise multiplication operation on specified tensor input operands.
cntk:element-xor Create an instance of the CNTK built-in elementwise XOR logical operation on the input operands.
cntk:elu Create an instance of the CNTK built-in elementwise exponential linear unit operation with the specified input operand.
cntk:elu-with-alpha Exponential linear unit operation.
cntk:embedding-layer Construct an embedding layer based and apply on operand.
cntk:end-static-axis Returns an axis representing the default end static axis.
cntk:equal Create an instance of the CNTK built-in elementwise equality comparison operation on specified tensor input operands.
cntk:evaluate Evaluate a CNTK model/function using specified input values.
cntk:exp Create an instance of the CNTK built-in elementwise exp operation on the input operands.
cntk:expand-dims Adds a singleton (size 1) axis at position axis.
cntk:eye-like Create an instance of a eye-like operation.
cntk:flatten Create an instance of a flatten operation that reshape the specified tensor into 2D tensor.
cntk:flatten-on-axis Create an instance of a flatten operation that reshape the specified tensor into 2D tensor.
cntk:floor Create an instance of the CNTK built-in elementwise floor operation on the input operands.
cntk:forward-backward Criterion node for training methods that rely on forward-backward Viterbi-like passes, e.g.
cntk:fs-ada-grad-learner Constructs an FSAdaGrad learner.
cntk:function Constructs a CNTK model/function from the binary node.
cntk:function-arguments Returns all arguments of a CNTK model, as a sequence of cntk:variable.
cntk:function-constants Returns the constants of a CNTK model, as a sequence of cntk:constant
cntk:function-from-variable Creates a pass-through function of the specified variable.
cntk:function-is-block Returns a boolean indicating if this Function is a block Function
cntk:function-is-composite Returns a boolean indicating if this Function is a composite Function
cntk:function-is-primitive Returns a boolean indicating if this Function is a primitive Function
cntk:function-name Returns the name of this function.
cntk:function-op-name Returns the name of the operation that this function denotes.
cntk:function-output Returns the output variable of a CNTK model.
cntk:function-outputs Returns the output variables of a CNTK model, as a sequence of cntk:variable
cntk:function-parameters Returns the parameters of a CNTK model, as a sequence of cntk:parameter
cntk:function-save Serializes a cntk:function as a node.
cntk:future-value This function returns the future value w.r.t.
cntk:gather-op Create an instance of the CNTK build-in operation to get a tensor that is gathered from reference tensor by indices.
cntk:gather-op-on-axis Create an instance of the CNTK build-in operation to get a tensor that is gathered from reference tensor by indices.
cntk:global-average-pooling-layer Layer factory function to create a global-average-pooling layer.
cntk:global-max-pooling-layer Layer factory function to create a global-max-pooling layer.
cntk:glorot-normal-initializer Returns a glorot normal parameter initializer.
cntk:glorot-uniform-initializer Returns a glorot uniform parameter initializer.
cntk:gpu Returns the GPU device based on provided CUDA device ID.
cntk:greater Create an instance of the CNTK built-in elementwise greater than comparison operation on specified tensor input operands.
cntk:greater-equal Create an instance of the CNTK built-in elementwise greater than or equal to comparison operation on specified tensor input operands.
cntk:gru-block Layer factory function to create a gru block.
cntk:gru-layer Layer factory function to create a GRU layer.
cntk:gumbel-random Create an instance of a Gumbel random operation.
cntk:gumbel-random-like Create an instance of a Gumbel random operation.
cntk:hard-sigmoid Create an instance of hard sigmoid operation: f(x) = max(0,min(alpha*x+beta,1))
cntk:hardmax Create an instance of the CNTK built-in elementwise hardmax operation on the input operands.
cntk:he-normal-initializer Returns a He normal parameter initializer.
cntk:he-uniform-initializer Returns a He uniform parameter initializer.
cntk:image-scaler Alteration of image by scaling its individual values.
cntk:input-variable Constructs an input variable.
cntk:label-layer Layer factory function to create a label layer.
cntk:labels-to-graph Conversion node from labels to graph.
cntk:lambda-rank Groups samples according to group, sorts them within each group based on output and computes the Normalized Discounted Cumulative Gain (NDCG) at infinity for each group.
cntk:lattice-sequence-with-softmax lattice-sequence-with-softmax.
cntk:layer-normalization-layer Layer factory function to create a layer-normalization layer.
cntk:leaky-re-lu Leaky Rectified linear operation.
cntk:learning-rate-schedule-from-constant Returns a learning rate schedule, where the learning rate is fixed throughout the whole training process.
cntk:learning-rate-schedule-from-pairs Returns a learning rate schedule, based on a sequence of pairs of "number of epochs" and "their learning rate".
cntk:learning-rate-schedule-from-sequence Returns a learning rate schedule, based on a sequence of learning rates.
cntk:less Create an instance of the CNTK built-in elementwise less than comparison operation on specified tensor input operands.
cntk:less-equal Create an instance of the CNTK built-in elementwise less than or equal to comparison operation on specified tensor input operands.
cntk:local-response-normalization Local Response Normalization layer.
cntk:log Create an instance of the CNTK built-in elementwise log operation on the input operands.
cntk:log-add-exp Create an instance of the CNTK built-in elementwise tensor operation that computes the log of the sum of the exponentials of the specified input operands.
cntk:log-softmax Create an instance of the CNTK built-in log softmax operation on a specified tensor input operand
cntk:log-softmax-on-axis Create an instance of the CNTK built-in log softmax operation on specified axis on a specified tensor input operand
cntk:lstm-block Layer factory function to create a lstm block.
cntk:lstm-layer Layer factory function to create a LSTM layer.
cntk:max-pooling-layer Layer factory function to create a max-pooling layer.
cntk:max-unpooling-layer Layer factory function to create a global-average-pooling layer.
cntk:mean Create a new Function instance that computes element-wise mean of input tensors.
cntk:mean-variance-normalization Computes mean-variance normalization of the specified input operand.
cntk:minibatch-size-schedule-from-constant Returns a minibatch size schedule, where the minibatch size remains the same.
cntk:minibatch-size-schedule-from-pairs Returns a minibatch size schedule.
cntk:minibatch-size-schedule-from-sequence Returns a minibatch size schedule, based on a sequence of minibatch sizes.
cntk:minus Create an instance of the CNTK built-in elementwise tensor subtraction operation with the specified input operands.
cntk:momentum-schedule-from-constant Returns a momentum schedule where the momentum decay rate is fixed.
cntk:momentum-schedule-from-pairs Returns a momentum schedule.
cntk:momentum-schedule-from-sequence Returns a momentum schedule, based on a sequence of momentum decay rates.
cntk:momentum-sgd-learner Constructs an momentum SGD learner.
cntk:ndcgat1 Create an instance of the CNTK built-in operation for evaluating the NDCG at 1 metric
cntk:negate Create an instance of the CNTK built-in elementwise negate operation on the input operands.
cntk:nesterov-learner Constructs an Nesterov learner.
cntk:new-unique-dynamic-axis Creates an Axis object representing a new unique dynamic axis.
cntk:normal-initializer Returns a normal parameter initializer.
cntk:normal-random Create an instance of a uniform random operation.
cntk:normal-random-like Create an instance of a normal random operation.
cntk:not-equal Create an instance of the CNTK built-in elementwise not-equal comparison operation on specified tensor input operands.
cntk:one-hot-op Create one hot tensor based on the input tensor.
cntk:ones-like Create an instance of a ones-like operation.
cntk:operand-sequence-axis Returns an axis representing the sequence axis (ordered dynamic axis) of an operand whose dynamic axes have not yet been inferred/resolved (i.e.
cntk:optimized-rnnstack An RNN implementation that uses the primitives in cuDNN.
cntk:output-variable Constructs an output variable.
cntk:pad Pads a tensor according to the specified patterns.
cntk:parameter Constructs a parameter from a parameter initializer.
cntk:parameter-from-scalar Constructs a parameter from a scalar value.
cntk:parameter-from-value Constructs a parameter from a cntk:value that is not a sequence or batch.
cntk:parameter-record-value-update After setting the value of parameter, update the record.
cntk:parameter-set-value Sets the value of the parameter, using a cntk:value that is not a sequence or batch.
cntk:parameter-value Returns the value of the parameter.
cntk:past-value This function returns the past value w.r.t.
cntk:per-dim-mean-variance-normalize Computes per dimension mean-variance normalization of the specified input operand.
cntk:placeholder-variable Constructs a placeholder variable.
cntk:plus Create an instance of the CNTK built-in elementwise tensor addition operation with the specified input operands.
cntk:pooling The pooling operations compute a new tensor by selecting the maximum or average value in the pooling input.
cntk:pooling-with-explicit-padding The pooling operations compute a new tensor by selecting the maximum or average value in the pooling input.
cntk:pow Create an instance of the CNTK built-in elementwise tensor operation that computes the leftOperand raised to the power of the right operand.
cntk:pre-lu Create an instance of the CNTK built-in elementwise parametric rectified linear Unit operation with the specified input operand and learning parameter alpha.
cntk:previous-minibatch-evaluation-average Average evaluation function value of the previous minibatch.
cntk:previous-minibatch-loss-average Average loss function value of the previous minibatch.
cntk:previous-minibatch-sample-count Number of sample encountered in the previous minibatch.
cntk:random-initializer-with-rank Overrides $output-rank and $filter-rank specification in a random initializer constructed without an explicit rank specification.
cntk:random-sample Estimates inclusion frequencies for random sampling with or without replacement.
cntk:random-sample-inclusion-frequency For weighted sampling with the specified sample size (num_samples) this operation computes the expected number of occurrences of each class in the sampled set.
cntk:reciprocal Create an instance of the CNTK built-in elementwise reciprocal operation on the input operands.
cntk:reconcile-dynamic-axes Create a new Function instance which reconciles the dynamic axes of the specified tensor operands.
cntk:reduce-l1 Computes the L1 norm of the input tensor’s element along the provided axes.
cntk:reduce-l2 Computes the L2 norm of the input tensor’s element along the provided axes.
cntk:reduce-log-sum Create an instance of the CNTK built-in LogSum reduction operation on specified tensor input operand along the specified axis
cntk:reduce-log-sum-on-axes Create an instance of the CNTK built-in LogSum reduction operation on specified tensor input operand along the specified axis
cntk:reduce-max Create an instance of the CNTK built-in Max reduction operation on specified tensor input operand along the specified axis
cntk:reduce-max-on-axes Create an instance of the CNTK built-in Max reduction operation on specified tensor input operand along the specified axis
cntk:reduce-mean Computes the mean of the input tensor’s elements across a specified axis or a list of specified axes.
cntk:reduce-mean-on-axes Create an instance of the CNTK built-in Mean reduction operation on specified tensor input operand along the specified axis
cntk:reduce-min Create an instance of the CNTK built-in Min reduction operation on specified tensor input operand along the specified axis
cntk:reduce-min-on-axes Create an instance of the CNTK built-in Mean reduction operation on specified tensor input operand along the specified axis
cntk:reduce-prod Computes the min of the input tensor’s elements across the specified axis.
cntk:reduce-prod-on-axes Create an instance of the CNTK built-in Prod reduction operation on specified tensor input operand along the specified axis
cntk:reduce-sum Computes the sum of the input tensor’s elements across one axis or a list of axes.
cntk:reduce-sum-on-axes Computes the sum of the input tensor’s elements across one axis or a list of axes.
cntk:reduce-sum-square Computes the sum square of the input tensor’s element along the provided axes.
cntk:relu Create an instance of the CNTK built-in elementwise relu operation on the input operands.
cntk:reshape Reinterpret input samples as having different tensor dimensions One dimension may be specified as 0 and will be inferred.
cntk:rms-prop-learner Constructs an RMSProp learner.
cntk:rnn-step-block Layer factory function to create a rnn-step block.
cntk:rnn-step-layer Layer factory function to create a RNN Step layer.
cntk:ROI-pooling The ROI-pooling operation computes a new matrix by selecting the maximum (max pooling) value in the pooling input for each region of interest (ROI).
cntk:round Create an instance of the CNTK built-in elementwise round operation on the input operands.
cntk:scalar Constructs a scalar, which is a zero dimensional constant.
cntk:selu Scaled exponential linear unit operation.
cntk:sequence Constructs a CNTK sequence as a cntk:value type.
cntk:sequence-delay This function combines past_value() and future_value() into a single function.
cntk:sequence-first Returns the first element of its symbolic input sequence
cntk:sequence-gather Takes two sequences of the same length and returns a new sequence whose elements are those elements of sequence whose corresponding element in condition is True, preserving the ordering of sequence.
cntk:sequence-gather-with-factor Takes two sequences of the same length and returns a new sequence whose elements are those elements of sequence whose corresponding element in condition is True, preserving the ordering of sequence.
cntk:sequence-is-first Returns a symbolic sequence of booleans with the same length as ``seq``.
cntk:sequence-is-last Returns a symbolic sequence of booleans with the same length as ``seq``.
cntk:sequence-last Returns the last element of its symbolic input sequence
cntk:sequence-reduce-max Create an instance of the CNTK built-in max reduction operation on specified tensor input operand along the operands lone dynamic sequence axis
cntk:sequence-reduce-sum Computes the sum of the input sequence’s elements across the sequence axis.
cntk:sequence-scatter Performs the inverse of gather.
cntk:sequence-scatter-with-factor Performs the inverse of gather.
cntk:sequence-slice Slice the input sequence.
cntk:sequence-softmax Create an instance of the CNTK built-in softmax operation on specified tensor input operand along the operands lone dynamic sequence axis
cntk:sequence-unpack Create an instance of the CNTK built-in operator for unpacking the specified sequence operand along the most significant static axis [-1] and padding any gaps with the specified padding value.
cntk:sequence-where Given a symbolic sequence condition of boolean-like (1/0) values, it will return a new sequence containing the indices for which the values were true.
cntk:sgd-learner Constructs an SGD learner.
cntk:shape Returns a shape object recognized by CNTK.
cntk:shape-dimensions Returns the dimensions represented by a cntk:shape object, as a sequence of unsigned long interger.
cntk:shape-free-dimension Returns a placeholder value to use for an axis whose dimension is unbound and is only determined when actual data is bound to the variable.
cntk:shape-has-free-dimension Returns a boolean value indicating if the dimension size for any of the axes of this shape is free.
cntk:shape-has-inferred-dimension Returns a boolean value indicating if the dimension size for any of the axes of this shape is unknown/inferred.
cntk:shape-has-unbound-dimension Returns a boolean value indicating if the dimension size for any of the axes of this shape is free or inferred.
cntk:shape-inferred-dimension Returns a placeholder value to use for an axis whose dimension is unknown and is to be inferred by the system.
cntk:shape-is-scalar Returns true if the shape represents a scalar.
cntk:shape-is-unknown Returns true if the shape is an unknown shape.
cntk:shape-rank Returns the rank of the shape.
cntk:shape-total-size Returns the total size of the rectangular shape that this shape denotes.
cntk:sigmoid Create an instance of the CNTK built-in elementwise sigmoid operation on the input operands.
cntk:sin Create an instance of the CNTK built-in elementwise sin operation on the input operands.
cntk:sinh Create an instance of the CNTK built-in elementwise sinh operation on the input operands.
cntk:slice Create an instance of the slice operation on specified tensor input operand
cntk:slice-with-strides Create an instance of the slice operation on specified tensor input operand
cntk:softmax Create an instance of the CNTK built-in softmax operation on specified tensor input operand
cntk:softmax-on-axis Create an instance of the CNTK built-in softmax operation on specified axis on a specified tensor input operand
cntk:softplus Softplus operation.
cntk:softsign Computes the element-wise softsign of x.
cntk:space-to-depth Rearranges elements in the input tensor from the spatial dimensions to the depth dimension.
cntk:splice Concatenate the input tensors along an axis.
cntk:sqrt Create an instance of the CNTK built-in elementwise sqrt operation on the input operands.
cntk:square Create an instance of the CNTK built-in elementwise square operation on the input operands.
cntk:squared-error This operation computes the sum of the squared difference between elements in the two input matrices.
cntk:squeeze Removes axes whose size is 1.
cntk:squeeze-axes Create an instance of the squeeze operation on specified tensor input operand, for the specified axis
cntk:stop-gradient Outputs its input as it is and prevents any gradient contribution from its output to its input.
cntk:straight-through Element-wise binarization node using the straight through estimator
cntk:subshape Constructs and returns a new cntk:shape object that is a subshape of $shape.
cntk:sum Create a new Function instance that computes element-wise sum of input tensors.
cntk:tan Create an instance of the CNTK built-in elementwise tan operation on the input operands.
cntk:tanh Create an instance of the CNTK built-in elementwise tanh operation on the input operands.
cntk:times The output of this operation is the matrix product of the two input matrices.
cntk:to-batch Create an instance of attach dynamic axis operation that convert the input's first static axis to dynamic axis.
cntk:to-sequence Create an instance of the CNTK built-in operator for converting the specified tensor operand into a sequence
cntk:to-sequence-like Create an instance of the CNTK built-in operator for converting the specified tensor operand into a sequence.
cntk:to-sequence-with-lengths Create an instance of the CNTK built-in operator for converting the specified tensor operand into a sequence This overload allows specifying an additional operand containing the lengths of individual sequences
cntk:top-k Create an instance of the CNTK built-in top k operation over the first static axis on a specified tensor input operand
cntk:top-k-on-axis Create an instance of the CNTK built-in top k operation over the first static axis on a specified tensor input operand
cntk:total-number-of-sample-seen The total number of samples seen by this trainer.
cntk:train-minibatch Optimize model parameter using a minibatch of training data.
cntk:trainer Constructs a trainer, used for training a model.
cntk:trainer-get-model Returns the model being trained by this trainer.
cntk:transpose Create an instance of the CNTK built-in log softmax operation on specified axis on a specified tensor input operand
cntk:transpose-axes Create an instance of the CNTK built-in transpose dimensions operation on specified tensor input operand.
cntk:transpose-on-axes Permutes the axes of the tensor.
cntk:transpose-times The output of this operation is the product of the first (left) argument with the second (right) argument transposed.
cntk:truncated-normal-initializer Truncated normal initializer.
cntk:uniform-initializer Returns a uniform parameter initializer.
cntk:uniform-random Create an instance of a uniform random operation.
cntk:uniform-random-like Create an instance of a uniform random operation.
cntk:unknown-dynamic-axes Returns unknown dynamic axes.
cntk:unknown-shape Returns a placeholder representing an unknown shape.
cntk:unpack-batch Create an instance of detach dynamic axis operation that convert the input's last dynamic axis to static axis.
cntk:unpooling Unpools the operand using information from pooling_input.
cntk:value Constructs a cntk:value.
cntk:value-alias Creates a new value which is an alias of this value.
cntk:value-copy-from Copies the contents of the source value to this Value.
cntk:value-data-type Returns the data type of this value's contents.
cntk:value-deep-clone Creates a new Value with newly allocated storage on the same device as this value and copies this value's contents into the newly allocated Value.
cntk:value-device Returns the descriptor of the device that this value resides on
cntk:value-erase Erase the data of the value.
cntk:value-is-readonly Returns true if the value is readonly.
cntk:value-is-valid Returns whether this object has been invalidated (by another forward and/or backward pass)
cntk:value-shape Returns the shape of this value's contents.
cntk:value-to-array Returns a json:array object representing the actual numeric values inside the cntk:value.
cntk:variable-as-string Returns the string representation of the variable.
cntk:variable-current-value-time-stamp Returns the timestamp of current value held by the variable.
cntk:variable-data-type Returns the data type of the variable.
cntk:variable-dynamic-axes Returns dynamic axes of the variable, as a sequence.
cntk:variable-from-function Constructs an output variable aliasing the output of the specified function.
cntk:variable-has-batch-axis Returns if the variable has batch axis.
cntk:variable-has-sequence-axis Returns if the variable has sequence axis.
cntk:variable-is-constant Returns if the variable is a constant variable.
cntk:variable-is-initialized Returns if the variable is initialized.
cntk:variable-is-input Returns if the variable is an input variable.
cntk:variable-is-output Returns if the variable is an output variable.
cntk:variable-is-parameter Returns if the variable is a parameter variable.
cntk:variable-is-placeholder Returns if the variable is a placeholder variable.
cntk:variable-is-sparse Returns if the variable is sparse.
cntk:variable-kind Returns the variable-kind of the variable (input, output, etc).
cntk:variable-name Returns the name of the variable.
cntk:variable-needs-gradient Returns if the variable needs gradient.
cntk:variable-owner Returns the function of which the specified variable is an output.
cntk:variable-shape Returns the shape of the variable.
cntk:variable-uid Returns the uid of the variable, as a string.
cntk:variable-value Returns the value of the variable.
cntk:weighted-binary-cross-entropy This operation computes the weighted binary cross entropy (aka logistic loss) between the output and target.
cntk:xavier-initializer Returns a xavier parameter initializer.
cntk:zeros-like Create an instance of a zeros-like operation.