In this blog, you will be introduced to Artificial Neural Network (ANN) which is one of the classes of machine learning algorithms.
INTRODUCTION
ANNs are designed to simulate the human brain algorithmically by exploiting the data. A ANN is composed of an input layer having input neurons, one or many hidden layers having neurons that learn the important features from data, and an output layer having output neurons. Here, the input neurons act as independent variables and output neurons act as dependent variables.
To understand the artificial neural network, we first need to understand a simple neuron.
PERCEPTRON
Perceptron is a single neural network or a single layer having multiple neurons.
A figure of single neuron,
In this figure, the circle depicted is a neuron which receives multiple input features along with the weights (w) that are multiplied with the input features (x).
The neuron also receives the bias value that helps the activation function to `decide that whether the current neuron should be activated or not . If it surpasses a particular threshold then the neuron will be updated. The bias value also represents if their is any biasness in the neuron for a particular feature. In multilayer architecture, we also use dropout values along with bias that helps the activation function to know if the neuron should be activated or not. While doing so, there will be times when the neuron will not get updated or learn the feature and that will help in reducing the biasness.
Going back to the perceptron, it consists one output, defined by y, which holds the value like,
Here, f is the activation function, n is the number of features, and theta is the bias value. The output of the perceptron is taken as binary as it has only one output.
ARIFICIAL NEURAL NETWORK
As we are now introduced with the perceptron, we are now going to talk about artificial neural network. Artificial Neural Network is made of up many such perceptrons that allow it to perform complex tasks than perceptron.
The hidden layer is placed between the input and output layers, and the hidden layer has direct connection to its subsequent layer, that is, the output layer. Each neuron, in the layers, is able to perform simple task and as artificial neural networks are composed of many neurons they are able to perform complex tasks.
The hidden layer has its own bias term. In ANN, the output layer can have multiple outputs which is contrast to the simple perceptron which only have single output. In ANN, we can use different activation functions for hidden layer and output layer.
If there are more than one hidden layer, then the neural network is called DNN, that is, Deep Neural Network.
The weights and bias terms are updated after receiving the output, to minimize the error and biasness so that accuracy can be improved.
If you need implementation for any of the topics mentioned above or assignment help on any of its variants, feel free to contact us.
Comments