Neural Networks: Industry use cases

5 min readMar 4, 2021

Deep learning

Deep learning is the area of artificial intelligence and also a customized part of machine learning. Deep learning works on deep neural network and its fundamental block is neuron. It is completely inspired form us, humans. So, we can say “Artificial intelligence is the intelligence given to the machine by the humans”. As we have all connected biological neurons inside us we are expecting from the machines.As these neurons help brain to take decision (intelligence), similar expectations we have from th machines. Traditional machine learning approaches was quite slow, so we started to work on deep learning which is faster, easy and accurate.

Variation of deep learning

Standard Neural Network
CNN (Convolutional Neural Network)
RNN (Recurrent Neural Network)
Hybrid Neural Network

Neuron

A neuron is simply understood as a function. A function that takes the feature as an input and gives the output after some internal process. This output may be the input for the other neurons or just the final output. It need weight and bias too to generate the output, that it gets from some initializer functions. Weight is different for each input whereas the bias is only one for each neuron.

It is an interconnected structure of neurons. A neural network has more than one neuron and each neuron connects with each input feature. If there are another layer of neurons then the output of each neuron will be the input of all the neurons of the next layer. It shows a connection like mesh topology. At last, we get the output based on linear regression on classification. These layers are broadly classified into 3 layers.

Input layer: This layer contains all the features that can be taken as input. All this input is in the form of a vector. Eg. x1, x2, x3,…xn.
Hidden layer: This layer contains all the layers of neurons. It is called a hidden layer because we don’t know about its exact appearance and if we don’t have the code then the number of layers is hidden from the user.
Output layer: This is the layer where we get the output. the output may be continuous or based on labels(classification).

Important vocabularies

Backpropagation: When the random initializer doesn’t initialize the weight and bias perfectly (loss function tells how perfect the weight is), the command goes back to the initializer function from the neuron and this process of going back is called back propogation. The ground truth value from the training set helps in it.
Feedforward: Each time when the backpropagation is done, the newly initialized value of weight and bias is fed to the neuron for the further process and it is called feedforwarding.
Perceptron: It is a simple neural network model, used for the supervised learning of binary classifiers.
Feature selection: Reversely it is also called feature elimination. Any data may have any number of features. Choosing which feature is important and will help in training and prediction is called feature selection.
Supervised learning: In supervised learning, we also have the output in the training set. This real output is called the Ground Truth value.

Real use cases of Neural Network

There are many companies that are using Artificial Intelligence for their business. The base of AI is a neural network only, so we can say that neural network is creating a new world for artificial intelligence. Some companies use it for analytics purpose, some use is to build the product, some uses it for the security, some use it to create a user-friendly environment., defense system also use it, agriculture projects are implemented using neural network. So the neural network is almost everywhere and expanding rapidly. Here, I am discussing some of its use cases of the neural network at the industrial level:

TESLA: Tesla is an American electric vehicle and clean energy company. It is quite popular these days for self-driving cars. Tesla is using 8 cameras on it to cover its surroundings to locate the obstacles and identify them. Then these images go through neural networks. The neural network is trained using PyTorch. Sometimes the result of the neural network is changed into 3D images. Depth estimation is something we generally do on stereo cameras. Having 2 cameras helps estimate distances better. Tesla is doing this using neural networks with a regression on the depth. Tesla is using all the collected data from its vehicles on the ride and using it to modify the features.
Netflix: As the world’s leading Internet television network with over 160 million members in over 190 countries, our members enjoy hundreds of millions of hours of content per day, including original series, documentaries, and feature films. Its recommendation is based on machine learning and neural network. It observes the history and thumbnail the user likes to open and by applying a neural network it gives a similar type of recommendation.
GOOGLE: Google search engine chorme is quite popular. In 2015 Google applied a 30 layers seep artificial neural network. Because of this much layers very complex searches become possible. These layers constantly helping the search engine to improve more and more. After appling this layer the error rate had dropped from 23% to 8%, said by the company. This is proved by the Goolge that if search engine will be improved the associated business with it will a also be improved. Amazon also followed it and its sales increased by 29%. This idea helped a lot to the E-commerce company to improve their search engine so that customers can get perfect things.
CLARIFAI: It is a New York, based artificial intelligence company. This company works on the concept of Deep Learning. It understands visual content (images and videos) in a way that’s similar to how the human brain processes images to identify the content. It uses CNN (Convolutuional Neural Network) for the processing of the visual content. It believes all the business from E-commerce to real state or maybe content management can use deep learning. The company offers its solution via API, mobile SDK, and on-premise solutions.
SALESFORCE: It is an American cloud-based company. It uses Deep Learning in data science (analytics work). Cloud software maker Salesforce created a platform called Einstein to simplify artificial intelligence and improve customer experiences with smarter and more personalized service. It provides customer relationship management (CRM) service and also provides a complementary suite of enterprise applications focused on customer service, marketing automation, analytics, and application development.

References:

I hope, you liked to read it. THANK YOU!!!