Artificial Neural Network

Humans are always curious about the processing capabilities and algorithms of the brain. The brain is the most advanced thing know to humans. Biologically, the brain consists of around 100 billion neurons.

What is a neuron?

The neuron is the basic working unit of the brain, a specialized cell designed to transmit information to other nerve cells. The data in a neuron is given by electrical signals and the data is processed and output is given in the form of electric signals.

Humans wanted to replicate the workings of the brain in order to create Artificial Intelligence. The first thing though was to create an artificial neuron-like structure, and this was achieved in 1958 by Frank Rosenblatt. This was called the perceptron which was a mathematical function for a neuron.

Perceptron

This was the mathematical model presented of a perceptron. The input in(t) was similar to the dendrites of a biological neuron that takes input wherein xᵢ being one of those multiple dendrites whereas out(t) was similar to the axon terminals known as synapses.

The decision boundary for the perceptron (and many other linear classifiers) is z = ∑ (xᵢwi) + b =0, this can be also written in the form as Matrix notations as,

Next, the function is passed through a sigmoid function. A sigmoid function is a mathematical function having a characteristic “S”-shaped curve or sigmoid curve. This function results in a binary value for the function passed through it, making this model a binary classifier.

The perceptron models were build to output a binary value. Changing the function changes the state of the output.

Neural Network

Now, we have a basic idea of a neuron, we would like to create a network on perceptrons creating a neural network resembling the network on neurons in the brain.

Check out the diagram above, every single circle represents a single neuron. For each neuron, the lines coming towards the neurons are where the neuron gets its input from, whereas the lines propagating from the neuron are its output. Note that the sigmoid function is applied on the last layer of neurons only, the neurons prior to the output layer have some other function, mostly relu. The layers between the input layer and the output layer are known as hidden layers.

What do we want to need Neural Networks?

The world today is of big data, and traditional machine learning algorithms fail once the amount of data grows tremendously. This means, after a point in time even if we increase the input data the output accuracy does not change. This is one of the many reasons why we need neural networks. The algorithms which use neural networks are terms as Deep Learning Algorithms.

Neural networks are capable of learning on the own, figuring the algorithm they require to get the output on their own. This does not mean that we should expect Neural networks to get the best out of the worst data. The neural network is capable of selecting the important features by itself. The more the neuron the better the predictive capability of the model.

Case Study of Neural Networks

Tesla Bets Farm On Neural Network Based Autonomy With Impressive Presentation — Forbes

There is one aspect of Tesla where it is miles ahead of the competition. And that is in its use of data to build what might just be the world’s most sophisticated, cutting-edge neural network anywhere.

Tesla’s use of data, AI, and ML to build a neural network — a system of sensors, data, communications, CPUs, peripheral hardware, and software that collectively processes information and adapts and learns like a human — is where the company really shines.

The ultimate endgame, according to Bowers, is “to look across all the neural networks and all the cars and bring all that information together and ultimately output one source of truth for the world around us.”

unlike traditional machine learning models, the deep learning algorithms could be trained directly on images, speeches, videos, and a lot more. They do not have a constraint as the algorithm is tailored every time. Neural networks are used from voice recognition to self-driving cars. Tesla uses deep neural networks to detect roads, cars, objects, and people in video feeds from eight cameras installed around the vehicle.

To conclude, the artificial neural network paves the way for Artificial Intelligence. Thank you and signing off...

#vimaldaga #righteducation #educationredefine #rightmentor #worldrecordholder #linuxworld #makingindiafutureready #righeducation #arthbylw #neuralnetwork #deeplearning

Final Year Student, understanding the industrial approach and tools