In this article, we are going to see how to optimize networking using optimization algorithms in PyBrain using Python.
In the field of machine learning, Optimization algorithms are specifically used to reduce certain functions known as loss function/error function. By loss function, the optimization algorithm may result in reducing the difference between the actual and predicted output. Eventually, building the model more accurate for the task. This article focuses on optimizing networking using optimization algorithms in PyBrain. PyBrain provides the support of GA optimization algorithm in order to optimize a network.
GA optimization algorithm
Step 1: Construct a classification dataset.
Let us firstly create a classification dataset. In this example, we have taken OR dataset.
Python3
# Python program to create a classification dataset# importing libraryfrom pybrain.datasets.classification import ClassificationDataSet# Creating OR datasetorDataset = ClassificationDataSet(2)# Inserting sample to orDatasetorDataset.addSample([0., 0.], [0.])orDataset.addSample([0., 1.], [1.])orDataset.addSample([1., 0.], [1.])orDataset.addSample([1., 1.], [1.])# Set the target fieldorDataset.setField('class', [[0.],[1.],[1.],[1.]]) |
Step 2: Creating a network.
To create a network, PyBrain provides us with pybrain.tools.shortcuts. We can import buildNetwork shortcuts from it.
Python3
# Python program to create a networkfrom pybrain.tools.shortcuts import buildNetwork# Building a network # The network consists of two input layers,# four hidden layers and one output layermyNetwork = buildNetwork(2, 4, 1) |
Step 3: Applying GA optimization algorithm.
The GA has the following syntax:
GA(dataset, network, minimize = True / False)
Here,
- dataset: A dataset
- network: The created network
- minimize = “True”: For reducing error function
Python3
from pybrain.optimization.populationbased.ga import GA# GA optimization algorithmgaOptimization = GA(orDataset.evaluateModuleMSE, myNetwork, minimize=True) |
Step 4: Applying learn operation.
After that, we need to iterate using a loop and optimize the created gaOptimization using learn(0) operation.
Python3
# 100 iterations for learningfor i in range(100): myNetwork = gaOptimization.learn(0)[0]# Giving input to activate the networkprint(myNetwork.activate([0, 0]))print(myNetwork.activate([1, 0]))print(myNetwork.activate([0, 1]))print(myNetwork.activate([1, 1])) |
Below is the complete implementation:
Python3
# Python program to demonstrate how to# optimize a network using Optimization# algorithms in PyBrain# Importing libraryfrom pybrain.datasets.classification import ClassificationDataSetfrom pybrain.tools.shortcuts import buildNetworkfrom pybrain.optimization.populationbased.ga import GA# Creating OR datasetorDataset = ClassificationDataSet(2)# Inserting sample to orDatasetorDataset.addSample([0., 0.], [0.])orDataset.addSample([0., 1.], [1.])orDataset.addSample([1., 0.], [1.])orDataset.addSample([1., 1.], [1.])# Set the target fieldorDataset.setField('class', [[0.], [1.], [1.], [1.]])# Building a network# The network consists of two input layers,# four hidden layers and one output layermyNetwork = buildNetwork(2, 4, 1)# GA optimization algorithmgaOptimization = GA(orDataset.evaluateModuleMSE, myNetwork, minimize=True)# 100 iterations for learningfor i in range(100): myNetwork = gaOptimization.learn(0)[0]# By passing input optimize the networkprint(myNetwork.activate([0, 0]))print(myNetwork.activate([1, 0]))print(myNetwork.activate([0, 1]))print(myNetwork.activate([1, 1])) |
Output:
