layer_kernel_conv2d: Perform Kernel Convolution

In tpq/caress: Caress, a Gentler Introduction to Keras

Description

This function calls k_conv2d as a "lambda" layer.

Usage

layer_kernel_conv2d(layer, kernel, name = NULL)

Arguments

layer	The incoming layer.
kernel	A layer that represents a kernel.
name	A string. The prefix label for all layers.

Value

The result of k_conv2d.

Examples

library(keras)
library(caress)
set.seed(1)
index <- sample(1:nrow(iris))
x <- as.matrix(iris[index,1:4])
y <- to_categorical(as.numeric(iris[index,5])-1)
k_clear_session()
input <- from_input(x)
reshape <- input %>%
  layer_reshape(c(4,1,1))
weight <- input %>%
  layer_learnable_array(c(1, 1, 4)) %>%
  layer_lambda(f = function(x) k_transpose(x))
target <- layer_kernel_conv2d(reshape, weight) %>%
  layer_flatten() %>%
  to_output(y)
m <- prepare(input, target)
build(m, x, y, batch_size = 4, epochs = 100)

tpq/caress documentation built on March 11, 2021, 8:03 p.m.