layer_learnable_array: Initialize Learnable Layer

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

Description

This function creates a layer of learnable weights. It requires connection to an input layer, but does not use the input. Instead, the values of the layer are the manifestation of learnable weights.

Usage

layer_learnable_array(input, array_dim, name = NULL)

Arguments

input	The incoming layer.
array_dim	The dimensions of the learnable kernel.
name	A string. The prefix label for all layers.

Value

A layer of learnable weights.

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)
weight <- input %>%
  layer_learnable_array(4) %>%
  layer_reshape(c(4,1))
target <- layer_kernel_dot(input, 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.