NoxmainNetwork 1.7.2

A package to create, train and test neural networks with mutiple layers

NoxmainNetwork

NoxmainNetwork is a Python3 module to create, train and test neural networks with mutiple layers. To use this module you should know how a neural network works.

Content

• Installation
• How to use

Installation

To install the NoxmainNetwork package, just dowload it via pip. To do so execute this command in your Terminal/cmd:

pip3 install NoxmainNetwork

How to use

After installing the module via pip, you can import it to a python project. Now you can create a neural network by using the class NoxmainNetwork. This class has the parameters neurons, random=True, silent=False.

• neurons: list Contains the neurons of the neural network. Each item is a layer with [itemvalue] neurons. The neural network needs at least two layers and one neuron in each layer.
• random: bool If true (default), the weights will be random when creating the neural network. Otherwise, every weight will be 0.0.
• silent: bool If false (default), the NoxmainNetwork prints a message when loading or saving. Otherwise, it will not.

Example:

import NoxmainNetwork

nn = NoxmainNetwork.NoxmainNetwork([50, 30, 10])

---

You can train the neural network by using the function train. It has the parameters inputs, targets, learningrate.

• inputs: list Contains one training iteration data. The length of the list has to be the number of neurons in the first layer. The values have to be betweet 0 and 1.
• targets: list Contains the correct ansewrs for the inputs. The length of the list has to be the number of neurons in the last layer. The values have to be betweet 0 and 1.
• learningrate: float A smaller learningrate results a slower learning but more accurate results. a larger learningrate results faster learning but more inaccurate results. The value of the learningrate has to be between 0 and 1.

Example:

nn.train([0.05, 0.99, 0.95, 0.88, 0.98, 0.46, 0.59, 0.89, 0.54, 0.27, 0.93, 0.37, 0.43, 0.87, 0.59, 0.04, 0.63, 0.7, 0.86, 0.44, 0.9, 0.66, 0.56, 0.2, 0.75, 0.1, 0.07, 0.36, 0.78, 0.39, 0.9, 0.53, 0.66, 0.46, 0.18, 0.75, 0.43, 0.11, 0.1, 1.0, 0.52, 0.39, 0.84, 0.21, 0.9, 0.84, 0.13, 0.83, 0.31, 0.43],
         [0.04, 0.52, 0.33, 0.13, 0.8, 0.19, 0.29, 0.27, 0.69, 0.26],
         0.1)

---

After training the neural network, you can test it by using the funtion query. It has the parameter inputs and returns a list.

• inputs: list Contains one training iteration data. The length of the list has to be the number of neurons in the first layer. The values have to be betweet 0 and 1.
• return The coutput from the NoxmainNetwork.

Example:

outputs = nn.query([0.98, 0.58, 0.51, 0.48, 0.12, 0.36, 0.14, 0.61, 0.92, 0.69, 0.73, 0.82, 0.12, 0.88, 0.7, 1.0, 0.75, 0.19, 0.43, 0.45, 0.22, 0.55, 0.47, 0.29, 0.27, 0.06, 0.08, 0.36, 0.82, 0.83, 1.0, 0.03, 0.13, 0.89, 0.6, 0.1, 0.35, 0.32, 0.47, 0.12, 0.72, 0.54, 0.08, 0.06, 0.56, 0.07, 0.83, 0.82, 0.87, 0.76])

---

You can save a neural network by using the function save. If this file already exists, it will be overwritten. The function has the parameters name, hide=False.

• name: str The filename of the saved NoxmainNetwork. The file will be called [name].npy.
• hide: bool If true, a dot is appended to the front to hide the file. Is false by default.

Example:

nn.save("nn")

---

You can load a saved neural network by using the function load. By loading a neural network the current weights will be overwritten by the loaded ones. The function has the parameters name, hidden=False.

• name: str The filename of the saved NoxmainNetwork. It tries to load the file called [name].npy.
• hidden: bool If false (default), it tries to load a visible file. Otherwise, it tries to load a hidden file (beginning with a dot).

Example:

nn.load("nn")