The CNN_1D companion object provides factory methods for creating 1D convolutional neural networks.
Attributes
Members list
Value members
Concrete methods
Build a matrix using forecasting.ARX.buildMatrix and call `rescale.
Build a matrix using forecasting.ARX.buildMatrix and call `rescale.
Value parameters
- f
-
the activation function family for layers 1->2 (input to hidden)
- f1
-
the activation function family for layers 2->3 (hidden to output)
- fname
-
the feature/variable names (defaults to null)
- hparam
-
the hyper-parameters for the model/network
- nc
-
the width of the filters (size of cofilters)
- nf
-
the number of filters for this convolutional layer
- pool
-
the pooling window size (if pool > 1, pooling is applied)
- poolFun
-
the pooling function to apply (e.g., CoFilter_1D.pool for max pooling, or CoFilter_1D.pool_a for average pooling)
- xe
-
the input/data matrix of exogenous varaibles
- y
-
the output/response vector, where y_i = response for element i of vector x
Attributes
Create a CNN_1D with automatic rescaling from a data matrix and response matrix.
Create a CNN_1D with automatic rescaling from a data matrix and response matrix.
Value parameters
- f
-
the activation function family for layers 1->2 (input to hidden)
- f1
-
the activation function family for layers 2->3 (hidden to output)
- fname
-
the feature/variable names (defaults to null)
- hparam
-
the hyper-parameters for the model/network
- nc
-
the width of the filters (size of cofilters)
- nf
-
the number of filters for this convolutional layer
- pool
-
the pooling window size (if pool > 1, pooling is applied)
- poolFun
-
the pooling function to apply (e.g., CoFilter_1D.pool for max pooling, or CoFilter_1D.pool_a for average pooling)
- x
-
the input/data matrix with instances stored in rows
- y
-
the output/response matrix, where y_i = response for row i of matrix x
Attributes
Update the parameters: the weights in the convolutional filter c and the weights biases in the fully-connected layer b.
Update the parameters: the weights in the convolutional filter c and the weights biases in the fully-connected layer b.
Value parameters
- b
-
the fully-connectd layer parameters
- filt
-
the array of convolutional filters
- nz
-
the width of the feature map per filter (from valid convolution)
- pool
-
the pooling window size (if pool > 1, pooling is applied
- x_
-
the training/full data/input matrix
- δ0
-
the convolutional layer delta
- δ1
-
the fully-connectd layer delta
- η
-
the learning rate
- φ
-
the training/full response/output matrix