Packages
neural_net
1.0.0
NeuralNet is an A.I. library that allows for the construction and training of complex recurrent neural networks. Architectures such as LSTM or GRU can be specified in under 20 lines of code. Any neural network that can be built with the NeuralNet DSL can be trainined with automatically implemente...
Current section
Files
Jump to
Current section
Files
lib/grui.ex
defmodule GRUI do
@moduledoc false
# @moduledoc "GRUI is an experimental neural network design in development."
use NeuralNet
def template(inp, out \\ uid) do
update_gate = sigmoid [inp, previous(out)]
negated_update_gate = mult_const update_gate, -1
forgetting_gate = add_const negated_update_gate, 1
content_weights = GRU.template(inp)
prev_out_gate = sigmoid [inp, previous(out)]
gated_prev_out = mult [prev_out_gate, previous(out)]
update_candidate = tanh_given_weights [inp, gated_prev_out], content_weights
gated_update = mult [update_candidate, update_gate]
purged_output = mult [previous(out), forgetting_gate]
add [purged_output, gated_update], out
end
defp define(args) do
template(input, output)
def_vec(input, args.input_ids)
def_vec(output, args.output_ids)
end
end