Burri Akhil Reddy

Burri Akhil Reddy

Use-cases of Neural Networks:

Use-cases of Neural Networks:

What are Neural Networks?

Neural network try to simulate the human brain by recognizing the patterns in the data. As these are created by human hence they are called as Artificial Neural Networks(ANN). ANN's are developed to perform cognitive functions such as decision making, problem solving and machine learning.

Components of a Neural Network:

Neural Networks have three layers the input layer: data enters the Neural Network through this layer, the hidden layer where the ouput from the previous input layer is used to recognize the patterns in the data, and the output layer where the Neural Networks give the output for the input data. There will be lot of neurons in each layer these neurons are called as perceptrons.

What is a perceptron?

perceptron.png

A perceptron is a unit which takes the input from the previous layers and applies a mathematical function on it and provides the output. These mathematical functions that the perceptrons apply on the input data are called as activation function. Here are some of the activation functions:

  • Linear(Default):

    linear.gif line.png

  • Sigmoid:

    sigmoid.gif sigm.png

  • Tanh:

    tanh.gif tanh.png

  • ReLU:

    leakyRelu.gif relu.png

  • Leaky ReLU:

    LeakyRelu

lrelu.png

  • Softmax:

    softmax.gif

How Neural Networks become intelligent?

Neural Networks are trained on a dataset with the input given in the first layer and the required output given in the output layer. During the training we the neural networks iterate over the data several times to get the insights from the dataset. While training based on the input certain output is given by the neural networks this output is compared with the original ouput to get the loss based on which it is decided which weight must have what value, and again the data is feeded into the neural network until it reaches the optimum value. In this process of training the neural networks find the weights such that the loss in the output is minimum. In this way by getting the optimum weights for the dataset the Neural Networks become intelligent.

Usecase in Industry:

Here are some of the usecases where Neural Networks is used:

  • Marketing Strategies:

    recommendations.png Usually recommendations on entertainment in apps such as youtube, netflix, prime and even in ecommerce sites such as amazon, flipkart all these applications use Neural Networks to give the recommendations. These apps use its users personal data and history on the application to decide what should be recommended. In this way they are helping the customers by giving the relavant information for their user and also marketing their products,

  • Filtering of e-mail:

    mail.png Applications like gmail filters the mails into groups like social, promotions, spam even for this kind of filtering Neural Networks is used. This will help the users to separate out the important and relevant mails from the irrelavant e-mail.

  • Identification and diagnosis of a disease:

    medicine.png In the field of medicine there is lot of data in the form of scan reports and test results this data is compared to a healty person's data based on which it is decided whether a person is infected with a disease or not. Neural Networks are trained with the test reports as the inputs and whether the person is diagnosed with a certain disease or not. And finally when some patients test reports are given it is decided by the model if the patient is tested positive or not.

  • Trading of Stocks:

stockmarket.png Neural networks are also used for trading they help in predicting the prices of the at a certain point in the future which helps the investors by giving the insights about in which direction the market will move.

And there are many more usecases.

 
Share this