Luparnet

Luparnet is a bare bones machine learning module that creates a simple class "net" This new object is a neural network that can be trained on data. The purpose of this system is to avoid having to use more complicated high level and complicated libraries and to serve as an education tool.

Installation

To install luparnet one simply needs to possess the Python Package Installer and enter the following into the shell:

pip install luparnet

Training your first network.

import luparnet as ln
import numpy as np

#input data
x = np.array([[0],[1]])

#output data
y = np.array([[1],[0]])

#build the network
myFirstNet=ln.net(2,[1,2,1])

#train the network and print the final loss
print("Final Loss: %.8f"%(np.mean(np.abs(myFirstNet.train(30000,x,y,error=5000)))))

#predict for the value one
print(myFirstNet.predict(np.array([[1]]),string=True))

This code is all you need to run a net on your computer, granted that you have both numpy and Luparnet installed.

Methods

ln.net(layer,ds,fuction="sigmoid")

Constructs a neural network. layer is equal to the amount of layers in the network ds is an array and initializes the dimensions of the network. function is the activation function with the default function being the sigmoid function see the Activation Functions section for more information and for a list of the functions.

yourNet.train(run,indata,outdata,error=False)

Trains the network for run times. indata and outdata are the input data and output data arrays of type np.array() error by default is False but takes an integer argument that prints the average loss, time elapsed and epoch of the training data at steps that are multiples of error

yourNet.predict(data)

Makes a prediction for data of type np.array(). Data must be an np.array().

ln.fit(csv)

Method takes a string, csv as input, must be a file name of a csv. The format should be a single set of input data points seperated by commas on each line. Returns an np.array() that can be used as inputs for the other methods.

Data Dimensions

It is extremely important to find the right dimensions of your array and understand exactly what that means to the network.

Take the example in Training Your Network

#input data
x = np.array([[0],[1]])

#output data
y = np.array([[1],[0]])

#build the network
myFirstNet=ln.net(2,[1,2,1])

In this example the input data has dimensions of 1,2 which then become the first two numbers of the ds array. The last digit of the array comes from the number of columns of the output array.

For a final example we can take the final example:

#input data
x = np.array([[1,0,1,1],[1,0,0,0],[0,0,0,0],[0,1,0,1]])

#output data
y = np.array([[1,1],[1,0],[1,0],[1,1]])

The output dimensions are therefore [4,4,2]

Activation Functions

The activation function defines the output of a given neuron or node within the network.