Neural Network

A neural network made up of layers of neurons connected to each other to form a relationship allowing it to learn.

After cloning:

$ mix deps.get
$ mix compile

Usage

Run the trainer and see the network learn using OR GATE data

$ mix learn or

You should see output like this:

OR gate learning *********************************************
Epoch: 0   Error: 0.0978034950879825143
Epoch: 1000   Error: 0.0177645755625382047
Epoch: 2000   Error: 0.0065019384961036274
Epoch: 3000   Error: 0.0032527653252166144
Epoch: 4000   Error: 0.0019254900093371497
Epoch: 5000   Error: 0.0012646710040632755
Epoch: 6000   Error: 0.0008910514800247452
Epoch: 7000   Error: 0.0006602873040322224
Epoch: 8000   Error: 0.0005081961006147329
Epoch: 9000   Error: 0.0004028528701046857
Epoch: 9999   Error: 0.0003270377487769315
Epoch: 10000   Error: 0.0003269728572615501
**************************************************************

Run the trainer and see the network learn using IRIS FLOWER GATE data

$ mix learn iris_flower

You should see output like this:

IRIS_FLOWER gate learning *********************************************
Epoch: 0   Error: 0.0164425788515711185
Epoch: 1000   Error: 0.027344153205250403
Epoch: 2000   Error: 0.0265533867778006451
Epoch: 3000   Error: 0.0266624718167679346
Epoch: 4000   Error: 0.0268164947904966262
Epoch: 5000   Error: 0.026857493502782933
Epoch: 6000   Error: 0.026794287038049043
Epoch: 7000   Error: 0.0266556275054049274
Epoch: 8000   Error: 0.0264642981722699525
Epoch: 9000   Error: 0.0262360305030914023
Epoch: 9999   Error: 0.025981881761432242
Epoch: 10000   Error: 0.025981617016649871
**************************************************************

Valid options are: or, and, xor, nand, iris_flower

Run tests

$ mix test

Run Console

alias NeuralNetwork.{DataFactory, Network, LossFunction, Layer}
{:ok, network_pid} = Network.start_link([2, 1], %{activation: :relu})
data = DataFactory.gate_for("or")
Network.fit(network_pid, data, %{epochs: 10_000, log_freqs: 1000})
Network.predict(network_pid, [1,1])

Huge props

• Levi Thompson https://github.com/levithomason
• Dev Coop group for support http://www.meetup.com/dev-coop/

Installation

Available in Hex, the package can be installed as:

1. Add neural_network to your list of dependencies in mix.exs:

   def deps do [{:neural_network, "~> 0.1.4"}] end

2. Ensure neural_network is started before your application:

   def application do [applications: [:neural_network]] end